slicenumpy.go

   1 // Copyright (C) The Lightning Authors. All rights reserved.
   2 //
   3 // SPDX-License-Identifier: AGPL-3.0
   4
   5 package lightning
   6
   7 import (
   8         "bufio"
   9         "bytes"
  10         "encoding/gob"
  11         "encoding/json"
  12         "errors"
  13         "flag"
  14         "fmt"
  15         "io"
  16         "io/ioutil"
  17         "math"
  18         "net/http"
  19         _ "net/http/pprof"
  20         "os"
  21         "regexp"
  22         "runtime"
  23         "runtime/debug"
  24         "sort"
  25         "strconv"
  26         "strings"
  27         "sync/atomic"
  28         "unsafe"
  29
  30         "git.arvados.org/arvados.git/sdk/go/arvados"
  31         "github.com/arvados/lightning/hgvs"
  32         "github.com/james-bowman/nlp"
  33         "github.com/kshedden/gonpy"
  34         "github.com/sirupsen/logrus"
  35         log "github.com/sirupsen/logrus"
  36         "golang.org/x/crypto/blake2b"
  37         "gonum.org/v1/gonum/mat"
  38 )
  39
  40 const annotationMaxTileSpan = 100
  41
  42 type sliceNumpy struct {
  43         filter             filter
  44         threads            int
  45         chi2Cases          []bool
  46         chi2PValue         float64
  47         pvalueMinFrequency float64
  48         maxFrequency       float64
  49         pcaComponents      int
  50         minCoverage        int
  51         minCoverageAll     bool
  52         includeVariant1    bool
  53         debugTag           tagID
  54
  55         cgnames         []string
  56         samples         []sampleInfo
  57         trainingSet     []int // samples index => training set index, or -1 if not in training set
  58         trainingSetSize int
  59         pvalue          func(onehot []bool) float64
  60         pvalueCallCount int64
  61 }
  62
  63 func (cmd *sliceNumpy) RunCommand(prog string, args []string, stdin io.Reader, stdout, stderr io.Writer) int {
  64         err := cmd.run(prog, args, stdin, stdout, stderr)
  65         if err != nil {
  66                 fmt.Fprintf(stderr, "%s\n", err)
  67                 return 1
  68         }
  69         return 0
  70 }
  71
  72 func (cmd *sliceNumpy) run(prog string, args []string, stdin io.Reader, stdout, stderr io.Writer) error {
  73         flags := flag.NewFlagSet("", flag.ContinueOnError)
  74         flags.SetOutput(stderr)
  75         pprof := flags.String("pprof", "", "serve Go profile data at http://`[addr]:port`")
  76         runlocal := flags.Bool("local", false, "run on local host (default: run in an arvados container)")
  77         arvadosRAM := flags.Int("arvados-ram", 750000000000, "amount of memory to request for arvados container (`bytes`)")
  78         arvadosVCPUs := flags.Int("arvados-vcpus", 96, "number of VCPUs to request for arvados container")
  79         projectUUID := flags.String("project", "", "project `UUID` for output data")
  80         priority := flags.Int("priority", 500, "container request priority")
  81         preemptible := flags.Bool("preemptible", true, "request preemptible instance")
  82         inputDir := flags.String("input-dir", "./in", "input `directory`")
  83         outputDir := flags.String("output-dir", "./out", "output `directory`")
  84         ref := flags.String("ref", "", "reference name (if blank, choose last one that appears in input)")
  85         regionsFilename := flags.String("regions", "", "only output columns/annotations that intersect regions in specified bed `file`")
  86         expandRegions := flags.Int("expand-regions", 0, "expand specified regions by `N` base pairs on each side`")
  87         mergeOutput := flags.Bool("merge-output", false, "merge output into one matrix.npy and one matrix.annotations.csv")
  88         hgvsSingle := flags.Bool("single-hgvs-matrix", false, "also generate hgvs-based matrix")
  89         hgvsChunked := flags.Bool("chunked-hgvs-matrix", false, "also generate hgvs-based matrix per chromosome")
  90         onehotSingle := flags.Bool("single-onehot", false, "generate one-hot tile-based matrix")
  91         onehotChunked := flags.Bool("chunked-onehot", false, "generate one-hot tile-based matrix per input chunk")
  92         samplesFilename := flags.String("samples", "", "`samples.csv` file with training/validation and case/control groups (see 'lightning choose-samples')")
  93         caseControlOnly := flags.Bool("case-control-only", false, "drop samples that are not in case/control groups")
  94         onlyPCA := flags.Bool("pca", false, "run principal component analysis, write components to pca.npy and samples.csv")
  95         flags.IntVar(&cmd.pcaComponents, "pca-components", 4, "number of PCA components to compute / use in logistic regression")
  96         maxPCATiles := flags.Int("max-pca-tiles", 0, "maximum tiles to use as PCA input (filter, then drop every 2nd colum pair until below max)")
  97         debugTag := flags.Int("debug-tag", -1, "log debugging details about specified tag")
  98         flags.BoolVar(&cmd.minCoverageAll, "min-coverage-all", false, "apply -min-coverage filter based on all samples, not just training set")
  99         flags.IntVar(&cmd.threads, "threads", 16, "number of memory-hungry assembly threads, and number of VCPUs to request for arvados container")
 100         flags.Float64Var(&cmd.chi2PValue, "chi2-p-value", 1, "do Χ² test (or logistic regression if -samples file has PCA components) and omit columns with p-value above this threshold")
 101         flags.Float64Var(&cmd.pvalueMinFrequency, "pvalue-min-frequency", 0.01, "skip p-value calculation on tile variants below this frequency in the training set")
 102         flags.Float64Var(&cmd.maxFrequency, "max-frequency", 1, "do not output variants above this frequency in the training set")
 103         flags.BoolVar(&cmd.includeVariant1, "include-variant-1", false, "include most common variant when building one-hot matrix")
 104         cmd.filter.Flags(flags)
 105         err := flags.Parse(args)
 106         if err == flag.ErrHelp {
 107                 return nil
 108         } else if err != nil {
 109                 return err
 110         } else if flags.NArg() > 0 {
 111                 return fmt.Errorf("errant command line arguments after parsed flags: %v", flags.Args())
 112         }
 113
 114         if *pprof != "" {
 115                 go func() {
 116                         log.Println(http.ListenAndServe(*pprof, nil))
 117                 }()
 118         }
 119
 120         if cmd.chi2PValue != 1 && *samplesFilename == "" {
 121                 return fmt.Errorf("cannot use provided -chi2-p-value=%f because -samples= value is empty", cmd.chi2PValue)
 122         }
 123
 124         cmd.debugTag = tagID(*debugTag)
 125
 126         if !*runlocal {
 127                 runner := arvadosContainerRunner{
 128                         Name:        "lightning slice-numpy",
 129                         Client:      arvados.NewClientFromEnv(),
 130                         ProjectUUID: *projectUUID,
 131                         RAM:         int64(*arvadosRAM),
 132                         VCPUs:       *arvadosVCPUs,
 133                         Priority:    *priority,
 134                         KeepCache:   2,
 135                         APIAccess:   true,
 136                         Preemptible: *preemptible,
 137                 }
 138                 err = runner.TranslatePaths(inputDir, regionsFilename, samplesFilename)
 139                 if err != nil {
 140                         return err
 141                 }
 142                 runner.Args = []string{"slice-numpy", "-local=true",
 143                         "-pprof=:6060",
 144                         "-input-dir=" + *inputDir,
 145                         "-output-dir=/mnt/output",
 146                         "-threads=" + fmt.Sprintf("%d", cmd.threads),
 147                         "-regions=" + *regionsFilename,
 148                         "-expand-regions=" + fmt.Sprintf("%d", *expandRegions),
 149                         "-merge-output=" + fmt.Sprintf("%v", *mergeOutput),
 150                         "-single-hgvs-matrix=" + fmt.Sprintf("%v", *hgvsSingle),
 151                         "-chunked-hgvs-matrix=" + fmt.Sprintf("%v", *hgvsChunked),
 152                         "-single-onehot=" + fmt.Sprintf("%v", *onehotSingle),
 153                         "-chunked-onehot=" + fmt.Sprintf("%v", *onehotChunked),
 154                         "-samples=" + *samplesFilename,
 155                         "-case-control-only=" + fmt.Sprintf("%v", *caseControlOnly),
 156                         "-min-coverage-all=" + fmt.Sprintf("%v", cmd.minCoverageAll),
 157                         "-pca=" + fmt.Sprintf("%v", *onlyPCA),
 158                         "-pca-components=" + fmt.Sprintf("%d", cmd.pcaComponents),
 159                         "-max-pca-tiles=" + fmt.Sprintf("%d", *maxPCATiles),
 160                         "-chi2-p-value=" + fmt.Sprintf("%f", cmd.chi2PValue),
 161                         "-pvalue-min-frequency=" + fmt.Sprintf("%f", cmd.pvalueMinFrequency),
 162                         "-max-frequency=" + fmt.Sprintf("%f", cmd.maxFrequency),
 163                         "-include-variant-1=" + fmt.Sprintf("%v", cmd.includeVariant1),
 164                         "-debug-tag=" + fmt.Sprintf("%d", cmd.debugTag),
 165                 }
 166                 runner.Args = append(runner.Args, cmd.filter.Args()...)
 167                 var output string
 168                 output, err = runner.Run()
 169                 if err != nil {
 170                         return err
 171                 }
 172                 fmt.Fprintln(stdout, output)
 173                 return nil
 174         }
 175
 176         infiles, err := allFiles(*inputDir, matchGobFile)
 177         if err != nil {
 178                 return err
 179         }
 180         if len(infiles) == 0 {
 181                 err = fmt.Errorf("no input files found in %s", *inputDir)
 182                 return err
 183         }
 184         sort.Strings(infiles)
 185
 186         var refseq map[string][]tileLibRef
 187         var reftiledata = make(map[tileLibRef][]byte, 11000000)
 188         in0, err := open(infiles[0])
 189         if err != nil {
 190                 return err
 191         }
 192
 193         matchGenome, err := regexp.Compile(cmd.filter.MatchGenome)
 194         if err != nil {
 195                 err = fmt.Errorf("-match-genome: invalid regexp: %q", cmd.filter.MatchGenome)
 196                 return err
 197         }
 198
 199         if *samplesFilename != "" {
 200                 cmd.samples, err = loadSampleInfo(*samplesFilename)
 201                 if err != nil {
 202                         return err
 203                 }
 204         } else if *caseControlOnly {
 205                 return fmt.Errorf("-case-control-only does not make sense without -samples")
 206         }
 207
 208         cmd.cgnames = nil
 209         var tagset [][]byte
 210         err = DecodeLibrary(in0, strings.HasSuffix(infiles[0], ".gz"), func(ent *LibraryEntry) error {
 211                 if len(ent.TagSet) > 0 {
 212                         tagset = ent.TagSet
 213                 }
 214                 for _, cseq := range ent.CompactSequences {
 215                         if cseq.Name == *ref || *ref == "" {
 216                                 refseq = cseq.TileSequences
 217                         }
 218                 }
 219                 for _, cg := range ent.CompactGenomes {
 220                         if matchGenome.MatchString(cg.Name) {
 221                                 cmd.cgnames = append(cmd.cgnames, cg.Name)
 222                         }
 223                 }
 224                 for _, tv := range ent.TileVariants {
 225                         if tv.Ref {
 226                                 reftiledata[tileLibRef{tv.Tag, tv.Variant}] = tv.Sequence
 227                         }
 228                 }
 229                 return nil
 230         })
 231         if err != nil {
 232                 return err
 233         }
 234         in0.Close()
 235         if refseq == nil {
 236                 err = fmt.Errorf("%s: reference sequence not found", infiles[0])
 237                 return err
 238         }
 239         if len(tagset) == 0 {
 240                 err = fmt.Errorf("tagset not found")
 241                 return err
 242         }
 243
 244         taglib := &tagLibrary{}
 245         err = taglib.setTags(tagset)
 246         if err != nil {
 247                 return err
 248         }
 249         taglen := taglib.TagLen()
 250         sort.Strings(cmd.cgnames)
 251
 252         if len(cmd.cgnames) == 0 {
 253                 return fmt.Errorf("fatal: 0 matching samples in library, nothing to do")
 254         }
 255         cmd.trainingSet = make([]int, len(cmd.cgnames))
 256         if *samplesFilename == "" {
 257                 cmd.trainingSetSize = len(cmd.cgnames)
 258                 for i, name := range cmd.cgnames {
 259                         cmd.samples = append(cmd.samples, sampleInfo{
 260                                 id:         trimFilenameForLabel(name),
 261                                 isTraining: true,
 262                         })
 263                         cmd.trainingSet[i] = i
 264                 }
 265         } else if len(cmd.cgnames) != len(cmd.samples) {
 266                 return fmt.Errorf("mismatched sample list: %d samples in library, %d in %s", len(cmd.cgnames), len(cmd.samples), *samplesFilename)
 267         } else {
 268                 for i, name := range cmd.cgnames {
 269                         if s := trimFilenameForLabel(name); s != cmd.samples[i].id {
 270                                 return fmt.Errorf("mismatched sample list: sample %d is %q in library, %q in %s", i, s, cmd.samples[i].id, *samplesFilename)
 271                         }
 272                 }
 273                 if *caseControlOnly {
 274                         for i := 0; i < len(cmd.samples); i++ {
 275                                 if !cmd.samples[i].isTraining && !cmd.samples[i].isValidation {
 276                                         if i+1 < len(cmd.samples) {
 277                                                 copy(cmd.samples[i:], cmd.samples[i+1:])
 278                                                 copy(cmd.cgnames[i:], cmd.cgnames[i+1:])
 279                                         }
 280                                         cmd.samples = cmd.samples[:len(cmd.samples)-1]
 281                                         cmd.cgnames = cmd.cgnames[:len(cmd.cgnames)-1]
 282                                         i--
 283                                 }
 284                         }
 285                 }
 286                 cmd.chi2Cases = nil
 287                 cmd.trainingSetSize = 0
 288                 for i := range cmd.cgnames {
 289                         if cmd.samples[i].isTraining {
 290                                 cmd.trainingSet[i] = cmd.trainingSetSize
 291                                 cmd.trainingSetSize++
 292                                 cmd.chi2Cases = append(cmd.chi2Cases, cmd.samples[i].isCase)
 293                         } else {
 294                                 cmd.trainingSet[i] = -1
 295                         }
 296                 }
 297                 if cmd.pvalue == nil {
 298                         cmd.pvalue = func(onehot []bool) float64 {
 299                                 return pvalue(onehot, cmd.chi2Cases)
 300                         }
 301                 }
 302         }
 303
 304         if cmd.minCoverageAll {
 305                 cmd.minCoverage = len(cmd.cgnames)
 306         } else {
 307                 cmd.minCoverage = cmd.trainingSetSize
 308         }
 309         if cmd.filter.MinCoverage < 1 {
 310                 cmd.minCoverage = int(math.Ceil(cmd.filter.MinCoverage * float64(cmd.minCoverage)))
 311         }
 312
 313         if len(cmd.samples[0].pcaComponents) > 0 {
 314                 cmd.pvalue = glmPvalueFunc(cmd.samples, cmd.pcaComponents)
 315                 // Unfortunately, statsmodel/glm lib logs stuff to
 316                 // os.Stdout when it panics on an unsolvable
 317                 // problem. We recover() from the panic in glm.go, but
 318                 // we also need to commandeer os.Stdout to avoid
 319                 // producing large quantities of logs.
 320                 stdoutWas := os.Stdout
 321                 defer func() { os.Stdout = stdoutWas }()
 322                 os.Stdout, err = os.Open(os.DevNull)
 323                 if err != nil {
 324                         return err
 325                 }
 326         }
 327
 328         // cgnamemap[name]==true for samples that we are including in
 329         // output
 330         cgnamemap := map[string]bool{}
 331         for _, name := range cmd.cgnames {
 332                 cgnamemap[name] = true
 333         }
 334
 335         err = writeSampleInfo(cmd.samples, *outputDir)
 336         if err != nil {
 337                 return err
 338         }
 339
 340         log.Info("indexing reference tiles")
 341         type reftileinfo struct {
 342                 variant  tileVariantID
 343                 seqname  string // chr1
 344                 pos      int    // distance from start of chromosome to starttag
 345                 tiledata []byte // acgtggcaa...
 346                 excluded bool   // true if excluded by regions file
 347                 nexttag  tagID  // tagID of following tile (-1 for last tag of chromosome)
 348         }
 349         isdup := map[tagID]bool{}
 350         reftile := map[tagID]*reftileinfo{}
 351         for seqname, cseq := range refseq {
 352                 pos := 0
 353                 lastreftag := tagID(-1)
 354                 for _, libref := range cseq {
 355                         if cmd.filter.MaxTag >= 0 && libref.Tag > tagID(cmd.filter.MaxTag) {
 356                                 continue
 357                         }
 358                         tiledata := reftiledata[libref]
 359                         if len(tiledata) == 0 {
 360                                 err = fmt.Errorf("missing tiledata for tag %d variant %d in %s in ref", libref.Tag, libref.Variant, seqname)
 361                                 return err
 362                         }
 363                         foundthistag := false
 364                         taglib.FindAll(tiledata[:len(tiledata)-1], func(tagid tagID, offset, _ int) {
 365                                 if !foundthistag && tagid == libref.Tag {
 366                                         foundthistag = true
 367                                         return
 368                                 }
 369                                 if dupref, ok := reftile[tagid]; ok {
 370                                         log.Printf("dropping reference tile %+v from %s @ %d, tag not unique, also found inside %+v from %s @ %d", tileLibRef{Tag: tagid, Variant: dupref.variant}, dupref.seqname, dupref.pos, libref, seqname, pos+offset+1)
 371                                         delete(reftile, tagid)
 372                                 } else {
 373                                         log.Printf("found tag %d at offset %d inside tile variant %+v on %s @ %d", tagid, offset, libref, seqname, pos+offset+1)
 374                                 }
 375                                 isdup[tagid] = true
 376                         })
 377                         if isdup[libref.Tag] {
 378                                 log.Printf("dropping reference tile %+v from %s @ %d, tag not unique", libref, seqname, pos)
 379                         } else if reftile[libref.Tag] != nil {
 380                                 log.Printf("dropping reference tile %+v from %s @ %d, tag not unique", tileLibRef{Tag: libref.Tag, Variant: reftile[libref.Tag].variant}, reftile[libref.Tag].seqname, reftile[libref.Tag].pos)
 381                                 delete(reftile, libref.Tag)
 382                                 log.Printf("dropping reference tile %+v from %s @ %d, tag not unique", libref, seqname, pos)
 383                                 isdup[libref.Tag] = true
 384                         } else {
 385                                 reftile[libref.Tag] = &reftileinfo{
 386                                         seqname:  seqname,
 387                                         variant:  libref.Variant,
 388                                         tiledata: tiledata,
 389                                         pos:      pos,
 390                                         nexttag:  -1,
 391                                 }
 392                                 if lastreftag >= 0 {
 393                                         reftile[lastreftag].nexttag = libref.Tag
 394                                 }
 395                                 lastreftag = libref.Tag
 396                         }
 397                         pos += len(tiledata) - taglen
 398                 }
 399                 log.Printf("... %s done, len %d", seqname, pos+taglen)
 400         }
 401
 402         var mask *mask
 403         if *regionsFilename != "" {
 404                 log.Printf("loading regions from %s", *regionsFilename)
 405                 mask, err = makeMask(*regionsFilename, *expandRegions)
 406                 if err != nil {
 407                         return err
 408                 }
 409                 log.Printf("before applying mask, len(reftile) == %d", len(reftile))
 410                 log.Printf("deleting reftile entries for regions outside %d intervals", mask.Len())
 411                 for _, rt := range reftile {
 412                         if !mask.Check(strings.TrimPrefix(rt.seqname, "chr"), rt.pos, rt.pos+len(rt.tiledata)) {
 413                                 rt.excluded = true
 414                         }
 415                 }
 416                 log.Printf("after applying mask, len(reftile) == %d", len(reftile))
 417         }
 418
 419         type hgvsColSet map[hgvs.Variant][2][]int8
 420         encodeHGVS := throttle{Max: len(refseq)}
 421         encodeHGVSTodo := map[string]chan hgvsColSet{}
 422         tmpHGVSCols := map[string]*os.File{}
 423         if *hgvsChunked {
 424                 for seqname := range refseq {
 425                         var f *os.File
 426                         f, err = os.Create(*outputDir + "/tmp." + seqname + ".gob")
 427                         if err != nil {
 428                                 return err
 429                         }
 430                         defer os.Remove(f.Name())
 431                         bufw := bufio.NewWriterSize(f, 1<<24)
 432                         enc := gob.NewEncoder(bufw)
 433                         tmpHGVSCols[seqname] = f
 434                         todo := make(chan hgvsColSet, 128)
 435                         encodeHGVSTodo[seqname] = todo
 436                         encodeHGVS.Go(func() error {
 437                                 for colset := range todo {
 438                                         err := enc.Encode(colset)
 439                                         if err != nil {
 440                                                 encodeHGVS.Report(err)
 441                                                 for range todo {
 442                                                 }
 443                                                 return err
 444                                         }
 445                                 }
 446                                 return bufw.Flush()
 447                         })
 448                 }
 449         }
 450
 451         var toMerge [][]int16
 452         if *mergeOutput || *hgvsSingle {
 453                 toMerge = make([][]int16, len(infiles))
 454         }
 455         var onehotIndirect [][2][]uint32 // [chunkIndex][axis][index]
 456         var onehotChunkSize []uint32
 457         var onehotXrefs [][]onehotXref
 458         if *onehotSingle || *onlyPCA {
 459                 onehotIndirect = make([][2][]uint32, len(infiles))
 460                 onehotChunkSize = make([]uint32, len(infiles))
 461                 onehotXrefs = make([][]onehotXref, len(infiles))
 462         }
 463         chunkStartTag := make([]tagID, len(infiles))
 464
 465         throttleMem := throttle{Max: cmd.threads} // TODO: estimate using mem and data size
 466         throttleNumpyMem := throttle{Max: cmd.threads/2 + 1}
 467         log.Info("generating annotations and numpy matrix for each slice")
 468         var errSkip = errors.New("skip infile")
 469         var done int64
 470         for infileIdx, infile := range infiles {
 471                 infileIdx, infile := infileIdx, infile
 472                 throttleMem.Go(func() error {
 473                         seq := make(map[tagID][]TileVariant, 50000)
 474                         cgs := make(map[string]CompactGenome, len(cmd.cgnames))
 475                         f, err := open(infile)
 476                         if err != nil {
 477                                 return err
 478                         }
 479                         defer f.Close()
 480                         log.Infof("%04d: reading %s", infileIdx, infile)
 481                         err = DecodeLibrary(f, strings.HasSuffix(infile, ".gz"), func(ent *LibraryEntry) error {
 482                                 for _, tv := range ent.TileVariants {
 483                                         if tv.Ref {
 484                                                 continue
 485                                         }
 486                                         // Skip tile with no
 487                                         // corresponding ref tile, if
 488                                         // mask is in play (we can't
 489                                         // determine coordinates for
 490                                         // these)
 491                                         if mask != nil && reftile[tv.Tag] == nil {
 492                                                 continue
 493                                         }
 494                                         // Skip tile whose
 495                                         // corresponding ref tile is
 496                                         // outside target regions --
 497                                         // unless it's a potential
 498                                         // spanning tile.
 499                                         if mask != nil && reftile[tv.Tag].excluded &&
 500                                                 (int(tv.Tag+1) >= len(tagset) ||
 501                                                         (bytes.HasSuffix(tv.Sequence, tagset[tv.Tag+1]) && reftile[tv.Tag+1] != nil && !reftile[tv.Tag+1].excluded)) {
 502                                                 continue
 503                                         }
 504                                         if tv.Tag == cmd.debugTag {
 505                                                 log.Printf("infile %d %s tag %d variant %d hash %x", infileIdx, infile, tv.Tag, tv.Variant, tv.Blake2b[:3])
 506                                         }
 507                                         variants := seq[tv.Tag]
 508                                         if len(variants) == 0 {
 509                                                 variants = make([]TileVariant, 100)
 510                                         }
 511                                         for len(variants) <= int(tv.Variant) {
 512                                                 variants = append(variants, TileVariant{})
 513                                         }
 514                                         variants[int(tv.Variant)] = tv
 515                                         seq[tv.Tag] = variants
 516                                 }
 517                                 for _, cg := range ent.CompactGenomes {
 518                                         if cmd.filter.MaxTag >= 0 && cg.StartTag > tagID(cmd.filter.MaxTag) {
 519                                                 return errSkip
 520                                         }
 521                                         if !cgnamemap[cg.Name] {
 522                                                 continue
 523                                         }
 524                                         // pad to full slice size
 525                                         // to avoid out-of-bounds
 526                                         // checks later
 527                                         if sliceSize := 2 * int(cg.EndTag-cg.StartTag); len(cg.Variants) < sliceSize {
 528                                                 cg.Variants = append(cg.Variants, make([]tileVariantID, sliceSize-len(cg.Variants))...)
 529                                         }
 530                                         cgs[cg.Name] = cg
 531                                 }
 532                                 return nil
 533                         })
 534                         if err == errSkip {
 535                                 return nil
 536                         } else if err != nil {
 537                                 return fmt.Errorf("%04d: DecodeLibrary(%s): %w", infileIdx, infile, err)
 538                         }
 539                         tagstart := cgs[cmd.cgnames[0]].StartTag
 540                         tagend := cgs[cmd.cgnames[0]].EndTag
 541                         chunkStartTag[infileIdx] = tagstart
 542
 543                         // TODO: filters
 544
 545                         log.Infof("%04d: renumber/dedup variants for tags %d-%d", infileIdx, tagstart, tagend)
 546                         variantRemap := make([][]tileVariantID, tagend-tagstart)
 547                         throttleCPU := throttle{Max: runtime.GOMAXPROCS(0)}
 548                         for tag, variants := range seq {
 549                                 tag, variants := tag, variants
 550                                 throttleCPU.Go(func() error {
 551                                         alleleCoverage := 0
 552                                         count := make(map[[blake2b.Size256]byte]int, len(variants))
 553
 554                                         rt := reftile[tag]
 555                                         if rt != nil {
 556                                                 count[blake2b.Sum256(rt.tiledata)] = 0
 557                                         }
 558
 559                                         for cgidx, cgname := range cmd.cgnames {
 560                                                 if !cmd.minCoverageAll && !cmd.samples[cgidx].isTraining {
 561                                                         continue
 562                                                 }
 563                                                 cg := cgs[cgname]
 564                                                 idx := int(tag-tagstart) * 2
 565                                                 for allele := 0; allele < 2; allele++ {
 566                                                         v := cg.Variants[idx+allele]
 567                                                         if v > 0 && len(variants[v].Sequence) > 0 {
 568                                                                 count[variants[v].Blake2b]++
 569                                                                 alleleCoverage++
 570                                                         }
 571                                                         if v > 0 && tag == cmd.debugTag {
 572                                                                 log.Printf("tag %d cg %s allele %d tv %d hash %x count is now %d", tag, cgname, allele, v, variants[v].Blake2b[:3], count[variants[v].Blake2b])
 573                                                         }
 574                                                 }
 575                                         }
 576                                         if alleleCoverage < cmd.minCoverage*2 {
 577                                                 idx := int(tag-tagstart) * 2
 578                                                 for _, cg := range cgs {
 579                                                         cg.Variants[idx] = 0
 580                                                         cg.Variants[idx+1] = 0
 581                                                 }
 582                                                 if tag == cmd.debugTag {
 583                                                         log.Printf("tag %d alleleCoverage %d < min %d, sample data wiped", tag, alleleCoverage, cmd.minCoverage*2)
 584                                                 }
 585                                                 return nil
 586                                         }
 587
 588                                         // hash[i] will be the hash of
 589                                         // the variant(s) that should
 590                                         // be at rank i (0-based).
 591                                         hash := make([][blake2b.Size256]byte, 0, len(count))
 592                                         for b := range count {
 593                                                 hash = append(hash, b)
 594                                         }
 595                                         sort.Slice(hash, func(i, j int) bool {
 596                                                 bi, bj := &hash[i], &hash[j]
 597                                                 if ci, cj := count[*bi], count[*bj]; ci != cj {
 598                                                         return ci > cj
 599                                                 } else {
 600                                                         return bytes.Compare((*bi)[:], (*bj)[:]) < 0
 601                                                 }
 602                                         })
 603                                         // rank[b] will be the 1-based
 604                                         // new variant number for
 605                                         // variants whose hash is b.
 606                                         rank := make(map[[blake2b.Size256]byte]tileVariantID, len(hash))
 607                                         for i, h := range hash {
 608                                                 rank[h] = tileVariantID(i + 1)
 609                                         }
 610                                         if tag == cmd.debugTag {
 611                                                 for h, r := range rank {
 612                                                         log.Printf("tag %d rank(%x) = %v", tag, h[:3], r)
 613                                                 }
 614                                         }
 615                                         // remap[v] will be the new
 616                                         // variant number for original
 617                                         // variant number v.
 618                                         remap := make([]tileVariantID, len(variants))
 619                                         for i, tv := range variants {
 620                                                 remap[i] = rank[tv.Blake2b]
 621                                         }
 622                                         if tag == cmd.debugTag {
 623                                                 for in, out := range remap {
 624                                                         if out > 0 {
 625                                                                 log.Printf("tag %d remap %d => %d", tag, in, out)
 626                                                         }
 627                                                 }
 628                                         }
 629                                         variantRemap[tag-tagstart] = remap
 630                                         if rt != nil {
 631                                                 refrank := rank[blake2b.Sum256(rt.tiledata)]
 632                                                 if tag == cmd.debugTag {
 633                                                         log.Printf("tag %d reftile variant %d => %d", tag, rt.variant, refrank)
 634                                                 }
 635                                                 rt.variant = refrank
 636                                         }
 637                                         return nil
 638                                 })
 639                         }
 640                         throttleCPU.Wait()
 641
 642                         var onehotChunk [][]int8
 643                         var onehotXref []onehotXref
 644
 645                         var annotationsFilename string
 646                         if *onlyPCA {
 647                                 annotationsFilename = "/dev/null"
 648                         } else {
 649                                 annotationsFilename = fmt.Sprintf("%s/matrix.%04d.annotations.csv", *outputDir, infileIdx)
 650                                 log.Infof("%04d: writing %s", infileIdx, annotationsFilename)
 651                         }
 652                         annof, err := os.Create(annotationsFilename)
 653                         if err != nil {
 654                                 return err
 655                         }
 656                         annow := bufio.NewWriterSize(annof, 1<<20)
 657                         outcol := 0
 658                         for tag := tagstart; tag < tagend; tag++ {
 659                                 rt := reftile[tag]
 660                                 if rt == nil && mask != nil {
 661                                         // With no ref tile, we don't
 662                                         // have coordinates to say
 663                                         // this is in the desired
 664                                         // regions -- so it's not.
 665                                         // TODO: handle ref spanning
 666                                         // tile case.
 667                                         continue
 668                                 }
 669                                 if rt != nil && rt.excluded {
 670                                         // TODO: don't skip yet --
 671                                         // first check for spanning
 672                                         // tile variants that
 673                                         // intersect non-excluded ref
 674                                         // tiles.
 675                                         continue
 676                                 }
 677                                 if cmd.filter.MaxTag >= 0 && tag > tagID(cmd.filter.MaxTag) {
 678                                         break
 679                                 }
 680                                 remap := variantRemap[tag-tagstart]
 681                                 if remap == nil {
 682                                         // was not assigned above,
 683                                         // because minCoverage
 684                                         outcol++
 685                                         continue
 686                                 }
 687                                 maxv := tileVariantID(0)
 688                                 for _, v := range remap {
 689                                         if maxv < v {
 690                                                 maxv = v
 691                                         }
 692                                 }
 693                                 if *onehotChunked || *onehotSingle || *onlyPCA {
 694                                         onehot, xrefs := cmd.tv2homhet(cgs, maxv, remap, tag, tagstart, seq)
 695                                         if tag == cmd.debugTag {
 696                                                 log.WithFields(logrus.Fields{
 697                                                         "onehot": onehot,
 698                                                         "xrefs":  xrefs,
 699                                                 }).Info("tv2homhet()")
 700                                         }
 701                                         onehotChunk = append(onehotChunk, onehot...)
 702                                         onehotXref = append(onehotXref, xrefs...)
 703                                 }
 704                                 if *onlyPCA {
 705                                         outcol++
 706                                         continue
 707                                 }
 708                                 if rt == nil {
 709                                         // Reference does not use any
 710                                         // variant of this tile
 711                                         //
 712                                         // TODO: diff against the
 713                                         // relevant portion of the
 714                                         // ref's spanning tile
 715                                         outcol++
 716                                         continue
 717                                 }
 718                                 fmt.Fprintf(annow, "%d,%d,%d,=,%s,%d,,,\n", tag, outcol, rt.variant, rt.seqname, rt.pos)
 719                                 variants := seq[tag]
 720                                 reftilestr := strings.ToUpper(string(rt.tiledata))
 721
 722                                 done := make([]bool, maxv+1)
 723                                 variantDiffs := make([][]hgvs.Variant, maxv+1)
 724                                 for v, tv := range variants {
 725                                         v := remap[v]
 726                                         if v == 0 || v == rt.variant || done[v] {
 727                                                 continue
 728                                         } else {
 729                                                 done[v] = true
 730                                         }
 731                                         if len(tv.Sequence) < taglen {
 732                                                 continue
 733                                         }
 734                                         // if reftilestr doesn't end
 735                                         // in the same tag as tv,
 736                                         // extend reftilestr with
 737                                         // following ref tiles until
 738                                         // it does (up to an arbitrary
 739                                         // sanity-check limit)
 740                                         reftilestr := reftilestr
 741                                         endtagstr := strings.ToUpper(string(tv.Sequence[len(tv.Sequence)-taglen:]))
 742                                         for i, rt := 0, rt; i < annotationMaxTileSpan && !strings.HasSuffix(reftilestr, endtagstr) && rt.nexttag >= 0; i++ {
 743                                                 rt = reftile[rt.nexttag]
 744                                                 if rt == nil {
 745                                                         break
 746                                                 }
 747                                                 reftilestr += strings.ToUpper(string(rt.tiledata[taglen:]))
 748                                         }
 749                                         if mask != nil && !mask.Check(strings.TrimPrefix(rt.seqname, "chr"), rt.pos, rt.pos+len(reftilestr)) {
 750                                                 continue
 751                                         }
 752                                         if !strings.HasSuffix(reftilestr, endtagstr) {
 753                                                 fmt.Fprintf(annow, "%d,%d,%d,,%s,%d,,,\n", tag, outcol, v, rt.seqname, rt.pos)
 754                                                 continue
 755                                         }
 756                                         if lendiff := len(reftilestr) - len(tv.Sequence); lendiff < -1000 || lendiff > 1000 {
 757                                                 fmt.Fprintf(annow, "%d,%d,%d,,%s,%d,,,\n", tag, outcol, v, rt.seqname, rt.pos)
 758                                                 continue
 759                                         }
 760                                         diffs, _ := hgvs.Diff(reftilestr, strings.ToUpper(string(tv.Sequence)), 0)
 761                                         for i := range diffs {
 762                                                 diffs[i].Position += rt.pos
 763                                         }
 764                                         for _, diff := range diffs {
 765                                                 fmt.Fprintf(annow, "%d,%d,%d,%s:g.%s,%s,%d,%s,%s,%s\n", tag, outcol, v, rt.seqname, diff.String(), rt.seqname, diff.Position, diff.Ref, diff.New, diff.Left)
 766                                         }
 767                                         if *hgvsChunked {
 768                                                 variantDiffs[v] = diffs
 769                                         }
 770                                 }
 771                                 if *hgvsChunked {
 772                                         // We can now determine, for each HGVS
 773                                         // variant (diff) in this reftile
 774                                         // region, whether a given genome
 775                                         // phase/allele (1) has the variant, (0) has
 776                                         // =ref or a different variant in that
 777                                         // position, or (-1) is lacking
 778                                         // coverage / couldn't be diffed.
 779                                         hgvsCol := hgvsColSet{}
 780                                         for _, diffs := range variantDiffs {
 781                                                 for _, diff := range diffs {
 782                                                         if _, ok := hgvsCol[diff]; ok {
 783                                                                 continue
 784                                                         }
 785                                                         hgvsCol[diff] = [2][]int8{
 786                                                                 make([]int8, len(cmd.cgnames)),
 787                                                                 make([]int8, len(cmd.cgnames)),
 788                                                         }
 789                                                 }
 790                                         }
 791                                         for row, name := range cmd.cgnames {
 792                                                 variants := cgs[name].Variants[(tag-tagstart)*2:]
 793                                                 for ph := 0; ph < 2; ph++ {
 794                                                         v := variants[ph]
 795                                                         if int(v) >= len(remap) {
 796                                                                 v = 0
 797                                                         } else {
 798                                                                 v = remap[v]
 799                                                         }
 800                                                         if v == rt.variant {
 801                                                                 // hgvsCol[*][ph][row] is already 0
 802                                                         } else if len(variantDiffs[v]) == 0 {
 803                                                                 // lacking coverage / couldn't be diffed
 804                                                                 for _, col := range hgvsCol {
 805                                                                         col[ph][row] = -1
 806                                                                 }
 807                                                         } else {
 808                                                                 for _, diff := range variantDiffs[v] {
 809                                                                         hgvsCol[diff][ph][row] = 1
 810                                                                 }
 811                                                         }
 812                                                 }
 813                                         }
 814                                         for diff, colpair := range hgvsCol {
 815                                                 allele2homhet(colpair)
 816                                                 if !cmd.filterHGVScolpair(colpair) {
 817                                                         delete(hgvsCol, diff)
 818                                                 }
 819                                         }
 820                                         if len(hgvsCol) > 0 {
 821                                                 encodeHGVSTodo[rt.seqname] <- hgvsCol
 822                                         }
 823                                 }
 824                                 outcol++
 825                         }
 826                         err = annow.Flush()
 827                         if err != nil {
 828                                 return err
 829                         }
 830                         err = annof.Close()
 831                         if err != nil {
 832                                 return err
 833                         }
 834
 835                         if *onehotChunked {
 836                                 // transpose onehotChunk[col][row] to numpy[row*ncols+col]
 837                                 rows := len(cmd.cgnames)
 838                                 cols := len(onehotChunk)
 839                                 log.Infof("%04d: preparing onehot numpy (rows=%d, cols=%d, mem=%d)", infileIdx, rows, cols, rows*cols)
 840                                 throttleNumpyMem.Acquire()
 841                                 out := onehotcols2int8(onehotChunk)
 842                                 fnm := fmt.Sprintf("%s/onehot.%04d.npy", *outputDir, infileIdx)
 843                                 err = writeNumpyInt8(fnm, out, rows, cols)
 844                                 if err != nil {
 845                                         return err
 846                                 }
 847                                 fnm = fmt.Sprintf("%s/onehot-columns.%04d.npy", *outputDir, infileIdx)
 848                                 err = writeNumpyInt32(fnm, onehotXref2int32(onehotXref), 4, len(onehotXref))
 849                                 if err != nil {
 850                                         return err
 851                                 }
 852                                 debug.FreeOSMemory()
 853                                 throttleNumpyMem.Release()
 854                         }
 855                         if *onehotSingle || *onlyPCA {
 856                                 onehotIndirect[infileIdx] = onehotChunk2Indirect(onehotChunk)
 857                                 onehotChunkSize[infileIdx] = uint32(len(onehotChunk))
 858                                 onehotXrefs[infileIdx] = onehotXref
 859                                 n := len(onehotIndirect[infileIdx][0])
 860                                 log.Infof("%04d: keeping onehot coordinates in memory (n=%d, mem=%d)", infileIdx, n, n*8*2)
 861                         }
 862                         if !(*onehotSingle || *onehotChunked || *onlyPCA) || *mergeOutput || *hgvsSingle {
 863                                 log.Infof("%04d: preparing numpy (rows=%d, cols=%d)", infileIdx, len(cmd.cgnames), 2*outcol)
 864                                 throttleNumpyMem.Acquire()
 865                                 rows := len(cmd.cgnames)
 866                                 cols := 2 * outcol
 867                                 out := make([]int16, rows*cols)
 868                                 for row, name := range cmd.cgnames {
 869                                         outidx := row * cols
 870                                         for col, v := range cgs[name].Variants {
 871                                                 tag := tagstart + tagID(col/2)
 872                                                 if cmd.filter.MaxTag >= 0 && tag > tagID(cmd.filter.MaxTag) {
 873                                                         break
 874                                                 }
 875                                                 if rt := reftile[tag]; mask != nil && (rt == nil || rt.excluded) {
 876                                                         continue
 877                                                 }
 878                                                 if v == 0 {
 879                                                         out[outidx] = 0 // tag not found / spanning tile
 880                                                 } else if variants, ok := seq[tag]; ok && int(v) < len(variants) && len(variants[v].Sequence) > 0 {
 881                                                         out[outidx] = int16(variantRemap[tag-tagstart][v])
 882                                                 } else {
 883                                                         out[outidx] = -1 // low quality tile variant
 884                                                 }
 885                                                 if tag == cmd.debugTag {
 886                                                         log.Printf("tag %d row %d col %d outidx %d v %d out %d", tag, row, col, outidx, v, out[outidx])
 887                                                 }
 888                                                 outidx++
 889                                         }
 890                                 }
 891                                 seq = nil
 892                                 cgs = nil
 893                                 debug.FreeOSMemory()
 894                                 throttleNumpyMem.Release()
 895                                 if *mergeOutput || *hgvsSingle {
 896                                         log.Infof("%04d: matrix fragment %d rows x %d cols", infileIdx, rows, cols)
 897                                         toMerge[infileIdx] = out
 898                                 }
 899                                 if !*mergeOutput && !*onehotChunked && !*onehotSingle {
 900                                         fnm := fmt.Sprintf("%s/matrix.%04d.npy", *outputDir, infileIdx)
 901                                         err = writeNumpyInt16(fnm, out, rows, cols)
 902                                         if err != nil {
 903                                                 return err
 904                                         }
 905                                 }
 906                         }
 907                         debug.FreeOSMemory()
 908                         log.Infof("%s: done (%d/%d)", infile, int(atomic.AddInt64(&done, 1)), len(infiles))
 909                         return nil
 910                 })
 911         }
 912         if err = throttleMem.Wait(); err != nil {
 913                 return err
 914         }
 915
 916         if *hgvsChunked {
 917                 log.Info("flushing hgvsCols temp files")
 918                 for seqname := range refseq {
 919                         close(encodeHGVSTodo[seqname])
 920                 }
 921                 err = encodeHGVS.Wait()
 922                 if err != nil {
 923                         return err
 924                 }
 925                 for seqname := range refseq {
 926                         log.Infof("%s: reading hgvsCols from temp file", seqname)
 927                         f := tmpHGVSCols[seqname]
 928                         _, err = f.Seek(0, io.SeekStart)
 929                         if err != nil {
 930                                 return err
 931                         }
 932                         var hgvsCols hgvsColSet
 933                         dec := gob.NewDecoder(bufio.NewReaderSize(f, 1<<24))
 934                         for err == nil {
 935                                 err = dec.Decode(&hgvsCols)
 936                         }
 937                         if err != io.EOF {
 938                                 return err
 939                         }
 940                         log.Infof("%s: sorting %d hgvs variants", seqname, len(hgvsCols))
 941                         variants := make([]hgvs.Variant, 0, len(hgvsCols))
 942                         for v := range hgvsCols {
 943                                 variants = append(variants, v)
 944                         }
 945                         sort.Slice(variants, func(i, j int) bool {
 946                                 vi, vj := &variants[i], &variants[j]
 947                                 if vi.Position != vj.Position {
 948                                         return vi.Position < vj.Position
 949                                 } else if vi.Ref != vj.Ref {
 950                                         return vi.Ref < vj.Ref
 951                                 } else {
 952                                         return vi.New < vj.New
 953                                 }
 954                         })
 955                         rows := len(cmd.cgnames)
 956                         cols := len(variants) * 2
 957                         log.Infof("%s: building hgvs matrix (rows=%d, cols=%d, mem=%d)", seqname, rows, cols, rows*cols)
 958                         out := make([]int8, rows*cols)
 959                         for varIdx, variant := range variants {
 960                                 hgvsCols := hgvsCols[variant]
 961                                 for row := range cmd.cgnames {
 962                                         for ph := 0; ph < 2; ph++ {
 963                                                 out[row*cols+varIdx+ph] = hgvsCols[ph][row]
 964                                         }
 965                                 }
 966                         }
 967                         err = writeNumpyInt8(fmt.Sprintf("%s/hgvs.%s.npy", *outputDir, seqname), out, rows, cols)
 968                         if err != nil {
 969                                 return err
 970                         }
 971                         out = nil
 972
 973                         fnm := fmt.Sprintf("%s/hgvs.%s.annotations.csv", *outputDir, seqname)
 974                         log.Infof("%s: writing hgvs column labels to %s", seqname, fnm)
 975                         var hgvsLabels bytes.Buffer
 976                         for varIdx, variant := range variants {
 977                                 fmt.Fprintf(&hgvsLabels, "%d,%s:g.%s\n", varIdx, seqname, variant.String())
 978                         }
 979                         err = ioutil.WriteFile(fnm, hgvsLabels.Bytes(), 0666)
 980                         if err != nil {
 981                                 return err
 982                         }
 983                 }
 984         }
 985
 986         if *mergeOutput || *hgvsSingle {
 987                 var annow *bufio.Writer
 988                 var annof *os.File
 989                 if *mergeOutput {
 990                         annoFilename := fmt.Sprintf("%s/matrix.annotations.csv", *outputDir)
 991                         annof, err = os.Create(annoFilename)
 992                         if err != nil {
 993                                 return err
 994                         }
 995                         annow = bufio.NewWriterSize(annof, 1<<20)
 996                 }
 997
 998                 rows := len(cmd.cgnames)
 999                 cols := 0
1000                 for _, chunk := range toMerge {
1001                         cols += len(chunk) / rows
1002                 }
1003                 log.Infof("merging output matrix (rows=%d, cols=%d, mem=%d) and annotations", rows, cols, rows*cols*2)
1004                 var out []int16
1005                 if *mergeOutput {
1006                         out = make([]int16, rows*cols)
1007                 }
1008                 hgvsCols := map[string][2][]int16{} // hgvs -> [[g0,g1,g2,...], [g0,g1,g2,...]] (slice of genomes for each phase)
1009                 startcol := 0
1010                 for outIdx, chunk := range toMerge {
1011                         chunkcols := len(chunk) / rows
1012                         if *mergeOutput {
1013                                 for row := 0; row < rows; row++ {
1014                                         copy(out[row*cols+startcol:], chunk[row*chunkcols:(row+1)*chunkcols])
1015                                 }
1016                         }
1017                         toMerge[outIdx] = nil
1018
1019                         annotationsFilename := fmt.Sprintf("%s/matrix.%04d.annotations.csv", *outputDir, outIdx)
1020                         log.Infof("reading %s", annotationsFilename)
1021                         buf, err := os.ReadFile(annotationsFilename)
1022                         if err != nil {
1023                                 return err
1024                         }
1025                         if *mergeOutput {
1026                                 err = os.Remove(annotationsFilename)
1027                                 if err != nil {
1028                                         return err
1029                                 }
1030                         }
1031                         for _, line := range bytes.Split(buf, []byte{'\n'}) {
1032                                 if len(line) == 0 {
1033                                         continue
1034                                 }
1035                                 fields := bytes.SplitN(line, []byte{','}, 9)
1036                                 tag, _ := strconv.Atoi(string(fields[0]))
1037                                 incol, _ := strconv.Atoi(string(fields[1]))
1038                                 tileVariant, _ := strconv.Atoi(string(fields[2]))
1039                                 hgvsID := string(fields[3])
1040                                 seqname := string(fields[4])
1041                                 pos, _ := strconv.Atoi(string(fields[5]))
1042                                 refseq := fields[6]
1043                                 if hgvsID == "" {
1044                                         // Null entry for un-diffable
1045                                         // tile variant
1046                                         continue
1047                                 }
1048                                 if hgvsID == "=" {
1049                                         // Null entry for ref tile
1050                                         continue
1051                                 }
1052                                 if mask != nil && !mask.Check(strings.TrimPrefix(seqname, "chr"), pos, pos+len(refseq)) {
1053                                         // The tile intersects one of
1054                                         // the selected regions, but
1055                                         // this particular HGVS
1056                                         // variant does not.
1057                                         continue
1058                                 }
1059                                 hgvsColPair := hgvsCols[hgvsID]
1060                                 if hgvsColPair[0] == nil {
1061                                         // values in new columns start
1062                                         // out as -1 ("no data yet")
1063                                         // or 0 ("=ref") here, may
1064                                         // change to 1 ("hgvs variant
1065                                         // present") below, either on
1066                                         // this line or a future line.
1067                                         hgvsColPair = [2][]int16{make([]int16, len(cmd.cgnames)), make([]int16, len(cmd.cgnames))}
1068                                         rt, ok := reftile[tagID(tag)]
1069                                         if !ok {
1070                                                 err = fmt.Errorf("bug: seeing annotations for tag %d, but it has no reftile entry", tag)
1071                                                 return err
1072                                         }
1073                                         for ph := 0; ph < 2; ph++ {
1074                                                 for row := 0; row < rows; row++ {
1075                                                         v := chunk[row*chunkcols+incol*2+ph]
1076                                                         if tileVariantID(v) == rt.variant {
1077                                                                 hgvsColPair[ph][row] = 0
1078                                                         } else {
1079                                                                 hgvsColPair[ph][row] = -1
1080                                                         }
1081                                                 }
1082                                         }
1083                                         hgvsCols[hgvsID] = hgvsColPair
1084                                         if annow != nil {
1085                                                 hgvsref := hgvs.Variant{
1086                                                         Position: pos,
1087                                                         Ref:      string(refseq),
1088                                                         New:      string(refseq),
1089                                                 }
1090                                                 fmt.Fprintf(annow, "%d,%d,%d,%s:g.%s,%s,%d,%s,%s,%s\n", tag, incol+startcol/2, rt.variant, seqname, hgvsref.String(), seqname, pos, refseq, refseq, fields[8])
1091                                         }
1092                                 }
1093                                 if annow != nil {
1094                                         fmt.Fprintf(annow, "%d,%d,%d,%s,%s,%d,%s,%s,%s\n", tag, incol+startcol/2, tileVariant, hgvsID, seqname, pos, refseq, fields[7], fields[8])
1095                                 }
1096                                 for ph := 0; ph < 2; ph++ {
1097                                         for row := 0; row < rows; row++ {
1098                                                 v := chunk[row*chunkcols+incol*2+ph]
1099                                                 if int(v) == tileVariant {
1100                                                         hgvsColPair[ph][row] = 1
1101                                                 }
1102                                         }
1103                                 }
1104                         }
1105
1106                         startcol += chunkcols
1107                 }
1108                 if *mergeOutput {
1109                         err = annow.Flush()
1110                         if err != nil {
1111                                 return err
1112                         }
1113                         err = annof.Close()
1114                         if err != nil {
1115                                 return err
1116                         }
1117                         err = writeNumpyInt16(fmt.Sprintf("%s/matrix.npy", *outputDir), out, rows, cols)
1118                         if err != nil {
1119                                 return err
1120                         }
1121                 }
1122                 out = nil
1123
1124                 if *hgvsSingle {
1125                         cols = len(hgvsCols) * 2
1126                         log.Printf("building hgvs-based matrix: %d rows x %d cols", rows, cols)
1127                         out = make([]int16, rows*cols)
1128                         hgvsIDs := make([]string, 0, cols/2)
1129                         for hgvsID := range hgvsCols {
1130                                 hgvsIDs = append(hgvsIDs, hgvsID)
1131                         }
1132                         sort.Strings(hgvsIDs)
1133                         var hgvsLabels bytes.Buffer
1134                         for idx, hgvsID := range hgvsIDs {
1135                                 fmt.Fprintf(&hgvsLabels, "%d,%s\n", idx, hgvsID)
1136                                 for ph := 0; ph < 2; ph++ {
1137                                         hgvscol := hgvsCols[hgvsID][ph]
1138                                         for row, val := range hgvscol {
1139                                                 out[row*cols+idx*2+ph] = val
1140                                         }
1141                                 }
1142                         }
1143                         err = writeNumpyInt16(fmt.Sprintf("%s/hgvs.npy", *outputDir), out, rows, cols)
1144                         if err != nil {
1145                                 return err
1146                         }
1147
1148                         fnm := fmt.Sprintf("%s/hgvs.annotations.csv", *outputDir)
1149                         log.Printf("writing hgvs labels: %s", fnm)
1150                         err = ioutil.WriteFile(fnm, hgvsLabels.Bytes(), 0777)
1151                         if err != nil {
1152                                 return err
1153                         }
1154                 }
1155         }
1156         if *onehotSingle || *onlyPCA {
1157                 nzCount := 0
1158                 for _, part := range onehotIndirect {
1159                         nzCount += len(part[0])
1160                 }
1161                 onehot := make([]uint32, nzCount*2) // [r,r,r,...,c,c,c,...]
1162                 var xrefs []onehotXref
1163                 chunkOffset := uint32(0)
1164                 outcol := 0
1165                 for i, part := range onehotIndirect {
1166                         for i := range part[1] {
1167                                 part[1][i] += chunkOffset
1168                         }
1169                         copy(onehot[outcol:], part[0])
1170                         copy(onehot[outcol+nzCount:], part[1])
1171                         xrefs = append(xrefs, onehotXrefs[i]...)
1172
1173                         outcol += len(part[0])
1174                         chunkOffset += onehotChunkSize[i]
1175
1176                         part[0] = nil
1177                         part[1] = nil
1178                         onehotXrefs[i] = nil
1179                         debug.FreeOSMemory()
1180                 }
1181                 if *onehotSingle {
1182                         fnm := fmt.Sprintf("%s/onehot.npy", *outputDir)
1183                         err = writeNumpyUint32(fnm, onehot, 2, nzCount)
1184                         if err != nil {
1185                                 return err
1186                         }
1187                         fnm = fmt.Sprintf("%s/onehot-columns.npy", *outputDir)
1188                         err = writeNumpyInt32(fnm, onehotXref2int32(xrefs), 5, len(xrefs))
1189                         if err != nil {
1190                                 return err
1191                         }
1192                         fnm = fmt.Sprintf("%s/stats.json", *outputDir)
1193                         j, err := json.Marshal(map[string]interface{}{
1194                                 "pvalueCallCount": cmd.pvalueCallCount,
1195                         })
1196                         if err != nil {
1197                                 return err
1198                         }
1199                         err = os.WriteFile(fnm, j, 0777)
1200                         if err != nil {
1201                                 return err
1202                         }
1203                 }
1204                 if *onlyPCA {
1205                         cols := 0
1206                         for _, c := range onehot[nzCount:] {
1207                                 if int(c) >= cols {
1208                                         cols = int(c) + 1
1209                                 }
1210                         }
1211                         if cols == 0 {
1212                                 return fmt.Errorf("cannot do PCA: one-hot matrix is empty")
1213                         }
1214                         log.Printf("have %d one-hot cols", cols)
1215                         stride := 1
1216                         for *maxPCATiles > 0 && cols > *maxPCATiles*2 {
1217                                 cols = (cols + 1) / 2
1218                                 stride = stride * 2
1219                         }
1220                         if cols%2 == 1 {
1221                                 // we work with pairs of columns
1222                                 cols++
1223                         }
1224                         log.Printf("creating full matrix (%d rows) and training matrix (%d rows) with %d cols, stride %d", len(cmd.cgnames), cmd.trainingSetSize, cols, stride)
1225                         mtxFull := mat.NewDense(len(cmd.cgnames), cols, nil)
1226                         mtxTrain := mat.NewDense(cmd.trainingSetSize, cols, nil)
1227                         for i, c := range onehot[nzCount:] {
1228                                 if int(c/2)%stride == 0 {
1229                                         outcol := int(c/2)/stride*2 + int(c)%2
1230                                         mtxFull.Set(int(onehot[i]), outcol, 1)
1231                                         if trainRow := cmd.trainingSet[int(onehot[i])]; trainRow >= 0 {
1232                                                 mtxTrain.Set(trainRow, outcol, 1)
1233                                         }
1234                                 }
1235                         }
1236                         log.Print("fitting")
1237                         transformer := nlp.NewPCA(cmd.pcaComponents)
1238                         transformer.Fit(mtxTrain.T())
1239                         log.Printf("transforming")
1240                         pca, err := transformer.Transform(mtxFull.T())
1241                         if err != nil {
1242                                 return err
1243                         }
1244                         pca = pca.T()
1245                         outrows, outcols := pca.Dims()
1246                         log.Printf("copying result to numpy output array: %d rows, %d cols", outrows, outcols)
1247                         out := make([]float64, outrows*outcols)
1248                         for i := 0; i < outrows; i++ {
1249                                 for j := 0; j < outcols; j++ {
1250                                         out[i*outcols+j] = pca.At(i, j)
1251                                 }
1252                         }
1253                         fnm := fmt.Sprintf("%s/pca.npy", *outputDir)
1254                         log.Printf("writing numpy: %s", fnm)
1255                         output, err := os.OpenFile(fnm, os.O_CREATE|os.O_TRUNC|os.O_WRONLY, 0777)
1256                         if err != nil {
1257                                 return err
1258                         }
1259                         npw, err := gonpy.NewWriter(nopCloser{output})
1260                         if err != nil {
1261                                 return fmt.Errorf("gonpy.NewWriter: %w", err)
1262                         }
1263                         npw.Shape = []int{outrows, outcols}
1264                         err = npw.WriteFloat64(out)
1265                         if err != nil {
1266                                 return fmt.Errorf("WriteFloat64: %w", err)
1267                         }
1268                         err = output.Close()
1269                         if err != nil {
1270                                 return err
1271                         }
1272                         log.Print("done")
1273
1274                         log.Print("copying pca components to sampleInfo")
1275                         for i := range cmd.samples {
1276                                 cmd.samples[i].pcaComponents = make([]float64, outcols)
1277                                 for c := 0; c < outcols; c++ {
1278                                         cmd.samples[i].pcaComponents[c] = pca.At(i, c)
1279                                 }
1280                         }
1281                         log.Print("done")
1282
1283                         err = writeSampleInfo(cmd.samples, *outputDir)
1284                         if err != nil {
1285                                 return err
1286                         }
1287                 }
1288         }
1289         if !*mergeOutput && !*onehotChunked && !*onehotSingle && !*onlyPCA {
1290                 tagoffsetFilename := *outputDir + "/chunk-tag-offset.csv"
1291                 log.Infof("writing tag offsets to %s", tagoffsetFilename)
1292                 var f *os.File
1293                 f, err = os.Create(tagoffsetFilename)
1294                 if err != nil {
1295                         return err
1296                 }
1297                 defer f.Close()
1298                 for idx, offset := range chunkStartTag {
1299                         _, err = fmt.Fprintf(f, "%q,%d\n", fmt.Sprintf("matrix.%04d.npy", idx), offset)
1300                         if err != nil {
1301                                 err = fmt.Errorf("write %s: %w", tagoffsetFilename, err)
1302                                 return err
1303                         }
1304                 }
1305                 err = f.Close()
1306                 if err != nil {
1307                         err = fmt.Errorf("close %s: %w", tagoffsetFilename, err)
1308                         return err
1309                 }
1310         }
1311
1312         return nil
1313 }
1314
1315 type sampleInfo struct {
1316         id            string
1317         isCase        bool
1318         isControl     bool
1319         isTraining    bool
1320         isValidation  bool
1321         pcaComponents []float64
1322 }
1323
1324 // Read samples.csv file with case/control and training/validation
1325 // flags.
1326 func loadSampleInfo(samplesFilename string) ([]sampleInfo, error) {
1327         var si []sampleInfo
1328         f, err := open(samplesFilename)
1329         if err != nil {
1330                 return nil, err
1331         }
1332         buf, err := io.ReadAll(f)
1333         f.Close()
1334         if err != nil {
1335                 return nil, err
1336         }
1337         lineNum := 0
1338         for _, csv := range bytes.Split(buf, []byte{'\n'}) {
1339                 lineNum++
1340                 if len(csv) == 0 {
1341                         continue
1342                 }
1343                 split := strings.Split(string(csv), ",")
1344                 if len(split) < 4 {
1345                         return nil, fmt.Errorf("%d fields < 4 in %s line %d: %q", len(split), samplesFilename, lineNum, csv)
1346                 }
1347                 if split[0] == "Index" && split[1] == "SampleID" && split[2] == "CaseControl" && split[3] == "TrainingValidation" {
1348                         continue
1349                 }
1350                 idx, err := strconv.Atoi(split[0])
1351                 if err != nil {
1352                         if lineNum == 1 {
1353                                 return nil, fmt.Errorf("header does not look right: %q", csv)
1354                         }
1355                         return nil, fmt.Errorf("%s line %d: index: %s", samplesFilename, lineNum, err)
1356                 }
1357                 if idx != len(si) {
1358                         return nil, fmt.Errorf("%s line %d: index %d out of order", samplesFilename, lineNum, idx)
1359                 }
1360                 var pcaComponents []float64
1361                 if len(split) > 4 {
1362                         for _, s := range split[4:] {
1363                                 f, err := strconv.ParseFloat(s, 64)
1364                                 if err != nil {
1365                                         return nil, fmt.Errorf("%s line %d: cannot parse float %q: %s", samplesFilename, lineNum, s, err)
1366                                 }
1367                                 pcaComponents = append(pcaComponents, f)
1368                         }
1369                 }
1370                 si = append(si, sampleInfo{
1371                         id:            split[1],
1372                         isCase:        split[2] == "1",
1373                         isControl:     split[2] == "0",
1374                         isTraining:    split[3] == "1",
1375                         isValidation:  split[3] == "0" && len(split[2]) > 0, // fix errant 0s in input
1376                         pcaComponents: pcaComponents,
1377                 })
1378         }
1379         return si, nil
1380 }
1381
1382 func writeSampleInfo(samples []sampleInfo, outputDir string) error {
1383         fnm := outputDir + "/samples.csv"
1384         log.Infof("writing sample metadata to %s", fnm)
1385         f, err := os.Create(fnm)
1386         if err != nil {
1387                 return err
1388         }
1389         defer f.Close()
1390         pcaLabels := ""
1391         if len(samples) > 0 {
1392                 for i := range samples[0].pcaComponents {
1393                         pcaLabels += fmt.Sprintf(",PCA%d", i)
1394                 }
1395         }
1396         _, err = fmt.Fprintf(f, "Index,SampleID,CaseControl,TrainingValidation%s\n", pcaLabels)
1397         if err != nil {
1398                 return err
1399         }
1400         for i, si := range samples {
1401                 var cc, tv string
1402                 if si.isCase {
1403                         cc = "1"
1404                 } else if si.isControl {
1405                         cc = "0"
1406                 }
1407                 if si.isTraining {
1408                         tv = "1"
1409                 } else if si.isValidation {
1410                         tv = "0"
1411                 }
1412                 var pcavals string
1413                 for _, pcaval := range si.pcaComponents {
1414                         pcavals += fmt.Sprintf(",%f", pcaval)
1415                 }
1416                 _, err = fmt.Fprintf(f, "%d,%s,%s,%s%s\n", i, si.id, cc, tv, pcavals)
1417                 if err != nil {
1418                         return fmt.Errorf("write %s: %w", fnm, err)
1419                 }
1420         }
1421         err = f.Close()
1422         if err != nil {
1423                 return fmt.Errorf("close %s: %w", fnm, err)
1424         }
1425         log.Print("done")
1426         return nil
1427 }
1428
1429 func (cmd *sliceNumpy) filterHGVScolpair(colpair [2][]int8) bool {
1430         if cmd.chi2PValue >= 1 {
1431                 return true
1432         }
1433         col0 := make([]bool, 0, len(cmd.chi2Cases))
1434         col1 := make([]bool, 0, len(cmd.chi2Cases))
1435         cases := make([]bool, 0, len(cmd.chi2Cases))
1436         for i, c := range cmd.chi2Cases {
1437                 if colpair[0][i] < 0 {
1438                         continue
1439                 }
1440                 col0 = append(col0, colpair[0][i] != 0)
1441                 col1 = append(col1, colpair[1][i] != 0)
1442                 cases = append(cases, c)
1443         }
1444         return len(cases) >= cmd.minCoverage &&
1445                 (pvalue(col0, cases) <= cmd.chi2PValue || pvalue(col1, cases) <= cmd.chi2PValue)
1446 }
1447
1448 func writeNumpyUint32(fnm string, out []uint32, rows, cols int) error {
1449         output, err := os.Create(fnm)
1450         if err != nil {
1451                 return err
1452         }
1453         defer output.Close()
1454         bufw := bufio.NewWriterSize(output, 1<<26)
1455         npw, err := gonpy.NewWriter(nopCloser{bufw})
1456         if err != nil {
1457                 return err
1458         }
1459         log.WithFields(log.Fields{
1460                 "filename": fnm,
1461                 "rows":     rows,
1462                 "cols":     cols,
1463                 "bytes":    rows * cols * 4,
1464         }).Infof("writing numpy: %s", fnm)
1465         npw.Shape = []int{rows, cols}
1466         npw.WriteUint32(out)
1467         err = bufw.Flush()
1468         if err != nil {
1469                 return err
1470         }
1471         return output.Close()
1472 }
1473
1474 func writeNumpyInt32(fnm string, out []int32, rows, cols int) error {
1475         output, err := os.Create(fnm)
1476         if err != nil {
1477                 return err
1478         }
1479         defer output.Close()
1480         bufw := bufio.NewWriterSize(output, 1<<26)
1481         npw, err := gonpy.NewWriter(nopCloser{bufw})
1482         if err != nil {
1483                 return err
1484         }
1485         log.WithFields(log.Fields{
1486                 "filename": fnm,
1487                 "rows":     rows,
1488                 "cols":     cols,
1489                 "bytes":    rows * cols * 4,
1490         }).Infof("writing numpy: %s", fnm)
1491         npw.Shape = []int{rows, cols}
1492         npw.WriteInt32(out)
1493         err = bufw.Flush()
1494         if err != nil {
1495                 return err
1496         }
1497         return output.Close()
1498 }
1499
1500 func writeNumpyInt16(fnm string, out []int16, rows, cols int) error {
1501         output, err := os.Create(fnm)
1502         if err != nil {
1503                 return err
1504         }
1505         defer output.Close()
1506         bufw := bufio.NewWriterSize(output, 1<<26)
1507         npw, err := gonpy.NewWriter(nopCloser{bufw})
1508         if err != nil {
1509                 return err
1510         }
1511         log.WithFields(log.Fields{
1512                 "filename": fnm,
1513                 "rows":     rows,
1514                 "cols":     cols,
1515                 "bytes":    rows * cols * 2,
1516         }).Infof("writing numpy: %s", fnm)
1517         npw.Shape = []int{rows, cols}
1518         npw.WriteInt16(out)
1519         err = bufw.Flush()
1520         if err != nil {
1521                 return err
1522         }
1523         return output.Close()
1524 }
1525
1526 func writeNumpyInt8(fnm string, out []int8, rows, cols int) error {
1527         output, err := os.Create(fnm)
1528         if err != nil {
1529                 return err
1530         }
1531         defer output.Close()
1532         bufw := bufio.NewWriterSize(output, 1<<26)
1533         npw, err := gonpy.NewWriter(nopCloser{bufw})
1534         if err != nil {
1535                 return err
1536         }
1537         log.WithFields(log.Fields{
1538                 "filename": fnm,
1539                 "rows":     rows,
1540                 "cols":     cols,
1541                 "bytes":    rows * cols,
1542         }).Infof("writing numpy: %s", fnm)
1543         npw.Shape = []int{rows, cols}
1544         npw.WriteInt8(out)
1545         err = bufw.Flush()
1546         if err != nil {
1547                 return err
1548         }
1549         return output.Close()
1550 }
1551
1552 func allele2homhet(colpair [2][]int8) {
1553         a, b := colpair[0], colpair[1]
1554         for i, av := range a {
1555                 bv := b[i]
1556                 if av < 0 || bv < 0 {
1557                         // no-call
1558                         a[i], b[i] = -1, -1
1559                 } else if av > 0 && bv > 0 {
1560                         // hom
1561                         a[i], b[i] = 1, 0
1562                 } else if av > 0 || bv > 0 {
1563                         // het
1564                         a[i], b[i] = 0, 1
1565                 } else {
1566                         // ref (or a different variant in same position)
1567                         // (this is a no-op) a[i], b[i] = 0, 0
1568                 }
1569         }
1570 }
1571
1572 type onehotXref struct {
1573         tag     tagID
1574         variant tileVariantID
1575         hom     bool
1576         pvalue  float64
1577         maf     float64
1578 }
1579
1580 const onehotXrefSize = unsafe.Sizeof(onehotXref{})
1581
1582 // Build onehot matrix (m[tileVariantIndex][genome] == 0 or 1) for all
1583 // variants of a single tile/tag#.
1584 //
1585 // Return nil if no tile variant passes Χ² filter.
1586 func (cmd *sliceNumpy) tv2homhet(cgs map[string]CompactGenome, maxv tileVariantID, remap []tileVariantID, tag, chunkstarttag tagID, seq map[tagID][]TileVariant) ([][]int8, []onehotXref) {
1587         if tag == cmd.debugTag {
1588                 tv := make([]tileVariantID, len(cmd.cgnames)*2)
1589                 for i, name := range cmd.cgnames {
1590                         copy(tv[i*2:(i+1)*2], cgs[name].Variants[(tag-chunkstarttag)*2:])
1591                 }
1592                 log.WithFields(logrus.Fields{
1593                         "cgs[i].Variants[tag*2+j]": tv,
1594                         "maxv":                     maxv,
1595                         "remap":                    remap,
1596                         "tag":                      tag,
1597                         "chunkstarttag":            chunkstarttag,
1598                 }).Info("tv2homhet()")
1599         }
1600         if maxv < 1 || (maxv < 2 && !cmd.includeVariant1) {
1601                 // everyone has the most common variant (of the variants we don't drop)
1602                 return nil, nil
1603         }
1604         tagoffset := tag - chunkstarttag
1605         coverage := 0
1606         for cgidx, cgname := range cmd.cgnames {
1607                 if !cmd.minCoverageAll && !cmd.samples[cgidx].isTraining {
1608                         continue
1609                 }
1610                 cg := cgs[cgname]
1611                 alleles := 0
1612                 for _, v := range cg.Variants[tagoffset*2 : tagoffset*2+2] {
1613                         if v > 0 && int(v) < len(seq[tag]) && len(seq[tag][v].Sequence) > 0 {
1614                                 alleles++
1615                         }
1616                 }
1617                 if alleles == 2 {
1618                         coverage++
1619                 }
1620         }
1621         if coverage < cmd.minCoverage {
1622                 return nil, nil
1623         }
1624         // "observed" array for p-value calculation (training set
1625         // only)
1626         obs := make([][]bool, (maxv+1)*2) // 2 slices (hom + het) for each variant#
1627         // one-hot output (all samples)
1628         outcols := make([][]int8, (maxv+1)*2)
1629         for i := range obs {
1630                 obs[i] = make([]bool, cmd.trainingSetSize)
1631                 outcols[i] = make([]int8, len(cmd.cgnames))
1632         }
1633         for cgid, name := range cmd.cgnames {
1634                 tsid := cmd.trainingSet[cgid]
1635                 cgvars := cgs[name].Variants[tagoffset*2:]
1636                 tv0, tv1 := remap[cgvars[0]], remap[cgvars[1]]
1637                 for v := tileVariantID(1); v <= maxv; v++ {
1638                         if tv0 == v && tv1 == v {
1639                                 if tsid >= 0 {
1640                                         obs[v*2][tsid] = true
1641                                 }
1642                                 outcols[v*2][cgid] = 1
1643                         } else if tv0 == v || tv1 == v {
1644                                 if tsid >= 0 {
1645                                         obs[v*2+1][tsid] = true
1646                                 }
1647                                 outcols[v*2+1][cgid] = 1
1648                         }
1649                 }
1650         }
1651         var onehot [][]int8
1652         var xref []onehotXref
1653         var maf float64
1654         for col := 2; col < len(obs); col++ {
1655                 // col 0,1 correspond to tile variant 0, i.e.,
1656                 // no-call; col 2,3 correspond to the most common
1657                 // variant; so we (normally) start at col 4.
1658                 if col < 4 && !cmd.includeVariant1 {
1659                         continue
1660                 }
1661                 if col&1 == 0 {
1662                         maf = homhet2maf(obs[col : col+2])
1663                         if maf < cmd.pvalueMinFrequency {
1664                                 // Skip both columns (hom and het) if
1665                                 // allele frequency is below threshold
1666                                 col++
1667                                 continue
1668                         }
1669                         if maf > cmd.maxFrequency {
1670                                 // Skip both columns if allele
1671                                 // frequency is above threshold
1672                                 col++
1673                                 continue
1674                         }
1675                 }
1676                 atomic.AddInt64(&cmd.pvalueCallCount, 1)
1677                 p := cmd.pvalue(obs[col])
1678                 if cmd.chi2PValue < 1 && !(p < cmd.chi2PValue) {
1679                         continue
1680                 }
1681                 onehot = append(onehot, outcols[col])
1682                 xref = append(xref, onehotXref{
1683                         tag:     tag,
1684                         variant: tileVariantID(col >> 1),
1685                         hom:     col&1 == 0,
1686                         pvalue:  p,
1687                         maf:     maf,
1688                 })
1689         }
1690         return onehot, xref
1691 }
1692
1693 func homhet2maf(onehot [][]bool) float64 {
1694         if len(onehot[0]) == 0 {
1695                 return 0
1696         }
1697         n := 0
1698         for i := range onehot[0] {
1699                 if onehot[0][i] {
1700                         // hom
1701                         n += 2
1702                 } else if onehot[1][i] {
1703                         // het
1704                         n += 1
1705                 }
1706         }
1707         return float64(n) / float64(len(onehot[0])*2)
1708 }
1709
1710 // convert a []onehotXref with length N to a numpy-style []int32
1711 // matrix with N columns, one row per field of onehotXref struct.
1712 //
1713 // Hom/het row contains hom=0, het=1.
1714 //
1715 // P-value row contains 1000000x actual p-value.
1716 func onehotXref2int32(xrefs []onehotXref) []int32 {
1717         xcols := len(xrefs)
1718         xdata := make([]int32, 6*xcols)
1719         for i, xref := range xrefs {
1720                 xdata[i] = int32(xref.tag)
1721                 xdata[xcols+i] = int32(xref.variant)
1722                 if xref.hom {
1723                         xdata[xcols*2+i] = 1
1724                 }
1725                 xdata[xcols*3+i] = int32(xref.pvalue * 1000000)
1726                 xdata[xcols*4+i] = int32(-math.Log10(xref.pvalue) * 1000000)
1727                 xdata[xcols*5+i] = int32(xref.maf * 1000000)
1728         }
1729         return xdata
1730 }
1731
1732 // transpose onehot data from in[col][row] to numpy-style
1733 // out[row*cols+col].
1734 func onehotcols2int8(in [][]int8) []int8 {
1735         if len(in) == 0 {
1736                 return nil
1737         }
1738         cols := len(in)
1739         rows := len(in[0])
1740         out := make([]int8, rows*cols)
1741         for row := 0; row < rows; row++ {
1742                 outrow := out[row*cols:]
1743                 for col, incol := range in {
1744                         outrow[col] = incol[row]
1745                 }
1746         }
1747         return out
1748 }
1749
1750 // Return [2][]uint32{rowIndices, colIndices} indicating which
1751 // elements of matrixT[c][r] have non-zero values.
1752 func onehotChunk2Indirect(matrixT [][]int8) [2][]uint32 {
1753         var nz [2][]uint32
1754         for c, col := range matrixT {
1755                 for r, val := range col {
1756                         if val != 0 {
1757                                 nz[0] = append(nz[0], uint32(r))
1758                                 nz[1] = append(nz[1], uint32(c))
1759                         }
1760                 }
1761         }
1762         return nz
1763 }