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         "errors"
  12         "flag"
  13         "fmt"
  14         "io"
  15         "io/ioutil"
  16         "math"
  17         "net/http"
  18         _ "net/http/pprof"
  19         "os"
  20         "regexp"
  21         "runtime"
  22         "runtime/debug"
  23         "sort"
  24         "strconv"
  25         "strings"
  26         "sync/atomic"
  27         "unsafe"
  28
  29         "git.arvados.org/arvados.git/sdk/go/arvados"
  30         "github.com/arvados/lightning/hgvs"
  31         "github.com/james-bowman/nlp"
  32         "github.com/kshedden/gonpy"
  33         "github.com/sirupsen/logrus"
  34         log "github.com/sirupsen/logrus"
  35         "golang.org/x/crypto/blake2b"
  36         "gonum.org/v1/gonum/mat"
  37 )
  38
  39 const annotationMaxTileSpan = 100
  40
  41 type sliceNumpy struct {
  42         filter          filter
  43         threads         int
  44         chi2Cases       []bool
  45         chi2PValue      float64
  46         minCoverage     int
  47         includeVariant1 bool
  48         debugTag        tagID
  49
  50         cgnames         []string
  51         samples         []sampleInfo
  52         trainingSet     []int // samples index => training set index, or -1 if not in training set
  53         trainingSetSize int
  54 }
  55
  56 func (cmd *sliceNumpy) RunCommand(prog string, args []string, stdin io.Reader, stdout, stderr io.Writer) int {
  57         err := cmd.run(prog, args, stdin, stdout, stderr)
  58         if err != nil {
  59                 fmt.Fprintf(stderr, "%s\n", err)
  60                 return 1
  61         }
  62         return 0
  63 }
  64
  65 func (cmd *sliceNumpy) run(prog string, args []string, stdin io.Reader, stdout, stderr io.Writer) error {
  66         flags := flag.NewFlagSet("", flag.ContinueOnError)
  67         flags.SetOutput(stderr)
  68         pprof := flags.String("pprof", "", "serve Go profile data at http://`[addr]:port`")
  69         runlocal := flags.Bool("local", false, "run on local host (default: run in an arvados container)")
  70         arvadosRAM := flags.Int("arvados-ram", 750000000000, "amount of memory to request for arvados container (`bytes`)")
  71         arvadosVCPUs := flags.Int("arvados-vcpus", 96, "number of VCPUs to request for arvados container")
  72         projectUUID := flags.String("project", "", "project `UUID` for output data")
  73         priority := flags.Int("priority", 500, "container request priority")
  74         inputDir := flags.String("input-dir", "./in", "input `directory`")
  75         outputDir := flags.String("output-dir", "./out", "output `directory`")
  76         ref := flags.String("ref", "", "reference name (if blank, choose last one that appears in input)")
  77         regionsFilename := flags.String("regions", "", "only output columns/annotations that intersect regions in specified bed `file`")
  78         expandRegions := flags.Int("expand-regions", 0, "expand specified regions by `N` base pairs on each side`")
  79         mergeOutput := flags.Bool("merge-output", false, "merge output into one matrix.npy and one matrix.annotations.csv")
  80         hgvsSingle := flags.Bool("single-hgvs-matrix", false, "also generate hgvs-based matrix")
  81         hgvsChunked := flags.Bool("chunked-hgvs-matrix", false, "also generate hgvs-based matrix per chromosome")
  82         onehotSingle := flags.Bool("single-onehot", false, "generate one-hot tile-based matrix")
  83         onehotChunked := flags.Bool("chunked-onehot", false, "generate one-hot tile-based matrix per input chunk")
  84         samplesFilename := flags.String("samples", "", "`samples.csv` file with training/validation and case/control groups (see 'lightning choose-samples')")
  85         caseControlOnly := flags.Bool("case-control-only", false, "drop samples that are not in case/control groups")
  86         onlyPCA := flags.Bool("pca", false, "run principal component analysis, write components to pca.npy and samples.csv")
  87         pcaComponents := flags.Int("pca-components", 4, "number of PCA components")
  88         maxPCATiles := flags.Int("max-pca-tiles", 0, "maximum tiles to use as PCA input (filter, then drop every 2nd colum pair until below max)")
  89         debugTag := flags.Int("debug-tag", -1, "log debugging details about specified tag")
  90         flags.IntVar(&cmd.threads, "threads", 16, "number of memory-hungry assembly threads, and number of VCPUs to request for arvados container")
  91         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")
  92         flags.BoolVar(&cmd.includeVariant1, "include-variant-1", false, "include most common variant when building one-hot matrix")
  93         cmd.filter.Flags(flags)
  94         err := flags.Parse(args)
  95         if err == flag.ErrHelp {
  96                 return nil
  97         } else if err != nil {
  98                 return err
  99         } else if flags.NArg() > 0 {
 100                 return fmt.Errorf("errant command line arguments after parsed flags: %v", flags.Args())
 101         }
 102
 103         if *pprof != "" {
 104                 go func() {
 105                         log.Println(http.ListenAndServe(*pprof, nil))
 106                 }()
 107         }
 108
 109         if cmd.chi2PValue != 1 && *samplesFilename == "" {
 110                 return fmt.Errorf("cannot use provided -chi2-p-value=%f because -samples= value is empty", cmd.chi2PValue)
 111         }
 112
 113         cmd.debugTag = tagID(*debugTag)
 114
 115         if !*runlocal {
 116                 runner := arvadosContainerRunner{
 117                         Name:        "lightning slice-numpy",
 118                         Client:      arvados.NewClientFromEnv(),
 119                         ProjectUUID: *projectUUID,
 120                         RAM:         int64(*arvadosRAM),
 121                         VCPUs:       *arvadosVCPUs,
 122                         Priority:    *priority,
 123                         KeepCache:   2,
 124                         APIAccess:   true,
 125                 }
 126                 err = runner.TranslatePaths(inputDir, regionsFilename, samplesFilename)
 127                 if err != nil {
 128                         return err
 129                 }
 130                 runner.Args = []string{"slice-numpy", "-local=true",
 131                         "-pprof=:6060",
 132                         "-input-dir=" + *inputDir,
 133                         "-output-dir=/mnt/output",
 134                         "-threads=" + fmt.Sprintf("%d", cmd.threads),
 135                         "-regions=" + *regionsFilename,
 136                         "-expand-regions=" + fmt.Sprintf("%d", *expandRegions),
 137                         "-merge-output=" + fmt.Sprintf("%v", *mergeOutput),
 138                         "-single-hgvs-matrix=" + fmt.Sprintf("%v", *hgvsSingle),
 139                         "-chunked-hgvs-matrix=" + fmt.Sprintf("%v", *hgvsChunked),
 140                         "-single-onehot=" + fmt.Sprintf("%v", *onehotSingle),
 141                         "-chunked-onehot=" + fmt.Sprintf("%v", *onehotChunked),
 142                         "-samples=" + *samplesFilename,
 143                         "-case-control-only=" + fmt.Sprintf("%v", *caseControlOnly),
 144                         "-pca=" + fmt.Sprintf("%v", *onlyPCA),
 145                         "-pca-components=" + fmt.Sprintf("%d", *pcaComponents),
 146                         "-max-pca-tiles=" + fmt.Sprintf("%d", *maxPCATiles),
 147                         "-chi2-p-value=" + fmt.Sprintf("%f", cmd.chi2PValue),
 148                         "-include-variant-1=" + fmt.Sprintf("%v", cmd.includeVariant1),
 149                         "-debug-tag=" + fmt.Sprintf("%d", cmd.debugTag),
 150                 }
 151                 runner.Args = append(runner.Args, cmd.filter.Args()...)
 152                 var output string
 153                 output, err = runner.Run()
 154                 if err != nil {
 155                         return err
 156                 }
 157                 fmt.Fprintln(stdout, output)
 158                 return nil
 159         }
 160
 161         infiles, err := allFiles(*inputDir, matchGobFile)
 162         if err != nil {
 163                 return err
 164         }
 165         if len(infiles) == 0 {
 166                 err = fmt.Errorf("no input files found in %s", *inputDir)
 167                 return err
 168         }
 169         sort.Strings(infiles)
 170
 171         var refseq map[string][]tileLibRef
 172         var reftiledata = make(map[tileLibRef][]byte, 11000000)
 173         in0, err := open(infiles[0])
 174         if err != nil {
 175                 return err
 176         }
 177
 178         matchGenome, err := regexp.Compile(cmd.filter.MatchGenome)
 179         if err != nil {
 180                 err = fmt.Errorf("-match-genome: invalid regexp: %q", cmd.filter.MatchGenome)
 181                 return err
 182         }
 183
 184         if *samplesFilename != "" {
 185                 cmd.samples, err = cmd.loadSampleInfo(*samplesFilename)
 186                 if err != nil {
 187                         return err
 188                 }
 189         } else if *caseControlOnly {
 190                 return fmt.Errorf("-case-control-only does not make sense without -samples")
 191         }
 192
 193         cmd.cgnames = nil
 194         var tagset [][]byte
 195         err = DecodeLibrary(in0, strings.HasSuffix(infiles[0], ".gz"), func(ent *LibraryEntry) error {
 196                 if len(ent.TagSet) > 0 {
 197                         tagset = ent.TagSet
 198                 }
 199                 for _, cseq := range ent.CompactSequences {
 200                         if cseq.Name == *ref || *ref == "" {
 201                                 refseq = cseq.TileSequences
 202                         }
 203                 }
 204                 for _, cg := range ent.CompactGenomes {
 205                         if matchGenome.MatchString(cg.Name) {
 206                                 cmd.cgnames = append(cmd.cgnames, cg.Name)
 207                         }
 208                 }
 209                 for _, tv := range ent.TileVariants {
 210                         if tv.Ref {
 211                                 reftiledata[tileLibRef{tv.Tag, tv.Variant}] = tv.Sequence
 212                         }
 213                 }
 214                 return nil
 215         })
 216         if err != nil {
 217                 return err
 218         }
 219         in0.Close()
 220         if refseq == nil {
 221                 err = fmt.Errorf("%s: reference sequence not found", infiles[0])
 222                 return err
 223         }
 224         if len(tagset) == 0 {
 225                 err = fmt.Errorf("tagset not found")
 226                 return err
 227         }
 228
 229         taglib := &tagLibrary{}
 230         err = taglib.setTags(tagset)
 231         if err != nil {
 232                 return err
 233         }
 234         taglen := taglib.TagLen()
 235         sort.Strings(cmd.cgnames)
 236
 237         if len(cmd.cgnames) == 0 {
 238                 return fmt.Errorf("fatal: 0 matching samples in library, nothing to do")
 239         }
 240         cmd.trainingSet = make([]int, len(cmd.cgnames))
 241         if *samplesFilename == "" {
 242                 cmd.trainingSetSize = len(cmd.cgnames)
 243                 for i, name := range cmd.cgnames {
 244                         cmd.samples = append(cmd.samples, sampleInfo{
 245                                 id:         trimFilenameForLabel(name),
 246                                 isTraining: true,
 247                         })
 248                         cmd.trainingSet[i] = i
 249                 }
 250         } else if len(cmd.cgnames) != len(cmd.samples) {
 251                 return fmt.Errorf("mismatched sample list: %d samples in library, %d in %s", len(cmd.cgnames), len(cmd.samples), *samplesFilename)
 252         } else {
 253                 for i, name := range cmd.cgnames {
 254                         if s := trimFilenameForLabel(name); s != cmd.samples[i].id {
 255                                 return fmt.Errorf("mismatched sample list: sample %d is %q in library, %q in %s", i, s, cmd.samples[i].id, *samplesFilename)
 256                         }
 257                 }
 258                 if *caseControlOnly {
 259                         for i := 0; i < len(cmd.samples); i++ {
 260                                 if !cmd.samples[i].isTraining && !cmd.samples[i].isValidation {
 261                                         if i+1 < len(cmd.samples) {
 262                                                 copy(cmd.samples[i:], cmd.samples[i+1:])
 263                                                 copy(cmd.cgnames[i:], cmd.cgnames[i+1:])
 264                                         }
 265                                         cmd.samples = cmd.samples[:len(cmd.samples)-1]
 266                                         cmd.cgnames = cmd.cgnames[:len(cmd.cgnames)-1]
 267                                         i--
 268                                 }
 269                         }
 270                 }
 271                 cmd.chi2Cases = nil
 272                 cmd.trainingSetSize = 0
 273                 for i := range cmd.cgnames {
 274                         if cmd.samples[i].isTraining {
 275                                 cmd.trainingSet[i] = cmd.trainingSetSize
 276                                 cmd.trainingSetSize++
 277                                 cmd.chi2Cases = append(cmd.chi2Cases, cmd.samples[i].isCase)
 278                         } else {
 279                                 cmd.trainingSet[i] = -1
 280                         }
 281                 }
 282         }
 283         if cmd.filter.MinCoverage == 1 {
 284                 // In the generic formula below, floating point
 285                 // arithmetic can effectively push the coverage
 286                 // threshold above 1.0, which is impossible/useless.
 287                 // 1.0 needs to mean exactly 100% coverage.
 288                 cmd.minCoverage = len(cmd.cgnames)
 289         } else {
 290                 cmd.minCoverage = int(math.Ceil(cmd.filter.MinCoverage * float64(len(cmd.cgnames))))
 291         }
 292
 293         {
 294                 samplesOutFilename := *outputDir + "/samples.csv"
 295                 log.Infof("writing sample metadata to %s", samplesOutFilename)
 296                 var f *os.File
 297                 f, err = os.Create(samplesOutFilename)
 298                 if err != nil {
 299                         return err
 300                 }
 301                 defer f.Close()
 302                 for i, si := range cmd.samples {
 303                         var cc, tv string
 304                         if si.isCase {
 305                                 cc = "1"
 306                         } else if si.isControl {
 307                                 cc = "0"
 308                         }
 309                         if si.isTraining {
 310                                 tv = "1"
 311                         } else {
 312                                 tv = "0"
 313                         }
 314                         _, err = fmt.Fprintf(f, "%d,%s,%s,%s\n", i, si.id, cc, tv)
 315                         if err != nil {
 316                                 err = fmt.Errorf("write %s: %w", samplesOutFilename, err)
 317                                 return err
 318                         }
 319                 }
 320                 err = f.Close()
 321                 if err != nil {
 322                         err = fmt.Errorf("close %s: %w", samplesOutFilename, err)
 323                         return err
 324                 }
 325                 log.Print("done")
 326         }
 327
 328         log.Info("indexing reference tiles")
 329         type reftileinfo struct {
 330                 variant  tileVariantID
 331                 seqname  string // chr1
 332                 pos      int    // distance from start of chromosome to starttag
 333                 tiledata []byte // acgtggcaa...
 334                 excluded bool   // true if excluded by regions file
 335                 nexttag  tagID  // tagID of following tile (-1 for last tag of chromosome)
 336         }
 337         isdup := map[tagID]bool{}
 338         reftile := map[tagID]*reftileinfo{}
 339         for seqname, cseq := range refseq {
 340                 pos := 0
 341                 lastreftag := tagID(-1)
 342                 for _, libref := range cseq {
 343                         if cmd.filter.MaxTag >= 0 && libref.Tag > tagID(cmd.filter.MaxTag) {
 344                                 continue
 345                         }
 346                         tiledata := reftiledata[libref]
 347                         if len(tiledata) == 0 {
 348                                 err = fmt.Errorf("missing tiledata for tag %d variant %d in %s in ref", libref.Tag, libref.Variant, seqname)
 349                                 return err
 350                         }
 351                         foundthistag := false
 352                         taglib.FindAll(tiledata[:len(tiledata)-1], func(tagid tagID, offset, _ int) {
 353                                 if !foundthistag && tagid == libref.Tag {
 354                                         foundthistag = true
 355                                         return
 356                                 }
 357                                 if dupref, ok := reftile[tagid]; ok {
 358                                         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)
 359                                         delete(reftile, tagid)
 360                                 } else {
 361                                         log.Printf("found tag %d at offset %d inside tile variant %+v on %s @ %d", tagid, offset, libref, seqname, pos+offset+1)
 362                                 }
 363                                 isdup[tagid] = true
 364                         })
 365                         if isdup[libref.Tag] {
 366                                 log.Printf("dropping reference tile %+v from %s @ %d, tag not unique", libref, seqname, pos)
 367                         } else if reftile[libref.Tag] != nil {
 368                                 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)
 369                                 delete(reftile, libref.Tag)
 370                                 log.Printf("dropping reference tile %+v from %s @ %d, tag not unique", libref, seqname, pos)
 371                                 isdup[libref.Tag] = true
 372                         } else {
 373                                 reftile[libref.Tag] = &reftileinfo{
 374                                         seqname:  seqname,
 375                                         variant:  libref.Variant,
 376                                         tiledata: tiledata,
 377                                         pos:      pos,
 378                                         nexttag:  -1,
 379                                 }
 380                                 if lastreftag >= 0 {
 381                                         reftile[lastreftag].nexttag = libref.Tag
 382                                 }
 383                                 lastreftag = libref.Tag
 384                         }
 385                         pos += len(tiledata) - taglen
 386                 }
 387                 log.Printf("... %s done, len %d", seqname, pos+taglen)
 388         }
 389
 390         var mask *mask
 391         if *regionsFilename != "" {
 392                 log.Printf("loading regions from %s", *regionsFilename)
 393                 mask, err = makeMask(*regionsFilename, *expandRegions)
 394                 if err != nil {
 395                         return err
 396                 }
 397                 log.Printf("before applying mask, len(reftile) == %d", len(reftile))
 398                 log.Printf("deleting reftile entries for regions outside %d intervals", mask.Len())
 399                 for _, rt := range reftile {
 400                         if !mask.Check(strings.TrimPrefix(rt.seqname, "chr"), rt.pos, rt.pos+len(rt.tiledata)) {
 401                                 rt.excluded = true
 402                         }
 403                 }
 404                 log.Printf("after applying mask, len(reftile) == %d", len(reftile))
 405         }
 406
 407         type hgvsColSet map[hgvs.Variant][2][]int8
 408         encodeHGVS := throttle{Max: len(refseq)}
 409         encodeHGVSTodo := map[string]chan hgvsColSet{}
 410         tmpHGVSCols := map[string]*os.File{}
 411         if *hgvsChunked {
 412                 for seqname := range refseq {
 413                         var f *os.File
 414                         f, err = os.Create(*outputDir + "/tmp." + seqname + ".gob")
 415                         if err != nil {
 416                                 return err
 417                         }
 418                         defer os.Remove(f.Name())
 419                         bufw := bufio.NewWriterSize(f, 1<<24)
 420                         enc := gob.NewEncoder(bufw)
 421                         tmpHGVSCols[seqname] = f
 422                         todo := make(chan hgvsColSet, 128)
 423                         encodeHGVSTodo[seqname] = todo
 424                         encodeHGVS.Go(func() error {
 425                                 for colset := range todo {
 426                                         err := enc.Encode(colset)
 427                                         if err != nil {
 428                                                 encodeHGVS.Report(err)
 429                                                 for range todo {
 430                                                 }
 431                                                 return err
 432                                         }
 433                                 }
 434                                 return bufw.Flush()
 435                         })
 436                 }
 437         }
 438
 439         var toMerge [][]int16
 440         if *mergeOutput || *hgvsSingle {
 441                 toMerge = make([][]int16, len(infiles))
 442         }
 443         var onehotIndirect [][2][]uint32 // [chunkIndex][axis][index]
 444         var onehotChunkSize []uint32
 445         var onehotXrefs [][]onehotXref
 446         if *onehotSingle || *onlyPCA {
 447                 onehotIndirect = make([][2][]uint32, len(infiles))
 448                 onehotChunkSize = make([]uint32, len(infiles))
 449                 onehotXrefs = make([][]onehotXref, len(infiles))
 450         }
 451         chunkStartTag := make([]tagID, len(infiles))
 452
 453         throttleMem := throttle{Max: cmd.threads} // TODO: estimate using mem and data size
 454         throttleNumpyMem := throttle{Max: cmd.threads/2 + 1}
 455         log.Info("generating annotations and numpy matrix for each slice")
 456         var errSkip = errors.New("skip infile")
 457         var done int64
 458         for infileIdx, infile := range infiles {
 459                 infileIdx, infile := infileIdx, infile
 460                 throttleMem.Go(func() error {
 461                         seq := make(map[tagID][]TileVariant, 50000)
 462                         cgs := make(map[string]CompactGenome, len(cmd.cgnames))
 463                         f, err := open(infile)
 464                         if err != nil {
 465                                 return err
 466                         }
 467                         defer f.Close()
 468                         log.Infof("%04d: reading %s", infileIdx, infile)
 469                         err = DecodeLibrary(f, strings.HasSuffix(infile, ".gz"), func(ent *LibraryEntry) error {
 470                                 for _, tv := range ent.TileVariants {
 471                                         if tv.Ref {
 472                                                 continue
 473                                         }
 474                                         // Skip tile with no
 475                                         // corresponding ref tile, if
 476                                         // mask is in play (we can't
 477                                         // determine coordinates for
 478                                         // these)
 479                                         if mask != nil && reftile[tv.Tag] == nil {
 480                                                 continue
 481                                         }
 482                                         // Skip tile whose
 483                                         // corresponding ref tile is
 484                                         // outside target regions --
 485                                         // unless it's a potential
 486                                         // spanning tile.
 487                                         if mask != nil && reftile[tv.Tag].excluded &&
 488                                                 (int(tv.Tag+1) >= len(tagset) ||
 489                                                         (bytes.HasSuffix(tv.Sequence, tagset[tv.Tag+1]) && reftile[tv.Tag+1] != nil && !reftile[tv.Tag+1].excluded)) {
 490                                                 continue
 491                                         }
 492                                         if tv.Tag == cmd.debugTag {
 493                                                 log.Printf("infile %d %s tag %d variant %d hash %x", infileIdx, infile, tv.Tag, tv.Variant, tv.Blake2b[:3])
 494                                         }
 495                                         variants := seq[tv.Tag]
 496                                         if len(variants) == 0 {
 497                                                 variants = make([]TileVariant, 100)
 498                                         }
 499                                         for len(variants) <= int(tv.Variant) {
 500                                                 variants = append(variants, TileVariant{})
 501                                         }
 502                                         variants[int(tv.Variant)] = tv
 503                                         seq[tv.Tag] = variants
 504                                 }
 505                                 for _, cg := range ent.CompactGenomes {
 506                                         if cmd.filter.MaxTag >= 0 && cg.StartTag > tagID(cmd.filter.MaxTag) {
 507                                                 return errSkip
 508                                         }
 509                                         if !matchGenome.MatchString(cg.Name) {
 510                                                 continue
 511                                         }
 512                                         // pad to full slice size
 513                                         // to avoid out-of-bounds
 514                                         // checks later
 515                                         if sliceSize := 2 * int(cg.EndTag-cg.StartTag); len(cg.Variants) < sliceSize {
 516                                                 cg.Variants = append(cg.Variants, make([]tileVariantID, sliceSize-len(cg.Variants))...)
 517                                         }
 518                                         cgs[cg.Name] = cg
 519                                 }
 520                                 return nil
 521                         })
 522                         if err == errSkip {
 523                                 return nil
 524                         } else if err != nil {
 525                                 return fmt.Errorf("%04d: DecodeLibrary(%s): err", infileIdx, infile)
 526                         }
 527                         tagstart := cgs[cmd.cgnames[0]].StartTag
 528                         tagend := cgs[cmd.cgnames[0]].EndTag
 529                         chunkStartTag[infileIdx] = tagstart
 530
 531                         // TODO: filters
 532
 533                         log.Infof("%04d: renumber/dedup variants for tags %d-%d", infileIdx, tagstart, tagend)
 534                         variantRemap := make([][]tileVariantID, tagend-tagstart)
 535                         throttleCPU := throttle{Max: runtime.GOMAXPROCS(0)}
 536                         for tag, variants := range seq {
 537                                 tag, variants := tag, variants
 538                                 throttleCPU.Go(func() error {
 539                                         alleleCoverage := 0
 540                                         count := make(map[[blake2b.Size256]byte]int, len(variants))
 541
 542                                         rt := reftile[tag]
 543                                         if rt != nil {
 544                                                 count[blake2b.Sum256(rt.tiledata)] = 0
 545                                         }
 546
 547                                         for cgname, cg := range cgs {
 548                                                 idx := int(tag-tagstart) * 2
 549                                                 for allele := 0; allele < 2; allele++ {
 550                                                         v := cg.Variants[idx+allele]
 551                                                         if v > 0 && len(variants[v].Sequence) > 0 {
 552                                                                 count[variants[v].Blake2b]++
 553                                                                 alleleCoverage++
 554                                                         }
 555                                                         if v > 0 && tag == cmd.debugTag {
 556                                                                 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])
 557                                                         }
 558                                                 }
 559                                         }
 560                                         if alleleCoverage < cmd.minCoverage*2 {
 561                                                 idx := int(tag-tagstart) * 2
 562                                                 for _, cg := range cgs {
 563                                                         cg.Variants[idx] = 0
 564                                                         cg.Variants[idx+1] = 0
 565                                                 }
 566                                                 if tag == cmd.debugTag {
 567                                                         log.Printf("tag %d alleleCoverage %d < min %d, sample data wiped", tag, alleleCoverage, cmd.minCoverage*2)
 568                                                 }
 569                                                 return nil
 570                                         }
 571
 572                                         // hash[i] will be the hash of
 573                                         // the variant(s) that should
 574                                         // be at rank i (0-based).
 575                                         hash := make([][blake2b.Size256]byte, 0, len(count))
 576                                         for b := range count {
 577                                                 hash = append(hash, b)
 578                                         }
 579                                         sort.Slice(hash, func(i, j int) bool {
 580                                                 bi, bj := &hash[i], &hash[j]
 581                                                 if ci, cj := count[*bi], count[*bj]; ci != cj {
 582                                                         return ci > cj
 583                                                 } else {
 584                                                         return bytes.Compare((*bi)[:], (*bj)[:]) < 0
 585                                                 }
 586                                         })
 587                                         // rank[b] will be the 1-based
 588                                         // new variant number for
 589                                         // variants whose hash is b.
 590                                         rank := make(map[[blake2b.Size256]byte]tileVariantID, len(hash))
 591                                         for i, h := range hash {
 592                                                 rank[h] = tileVariantID(i + 1)
 593                                         }
 594                                         if tag == cmd.debugTag {
 595                                                 for h, r := range rank {
 596                                                         log.Printf("tag %d rank(%x) = %v", tag, h[:3], r)
 597                                                 }
 598                                         }
 599                                         // remap[v] will be the new
 600                                         // variant number for original
 601                                         // variant number v.
 602                                         remap := make([]tileVariantID, len(variants))
 603                                         for i, tv := range variants {
 604                                                 remap[i] = rank[tv.Blake2b]
 605                                         }
 606                                         if tag == cmd.debugTag {
 607                                                 for in, out := range remap {
 608                                                         if out > 0 {
 609                                                                 log.Printf("tag %d remap %d => %d", tag, in, out)
 610                                                         }
 611                                                 }
 612                                         }
 613                                         variantRemap[tag-tagstart] = remap
 614                                         if rt != nil {
 615                                                 refrank := rank[blake2b.Sum256(rt.tiledata)]
 616                                                 if tag == cmd.debugTag {
 617                                                         log.Printf("tag %d reftile variant %d => %d", tag, rt.variant, refrank)
 618                                                 }
 619                                                 rt.variant = refrank
 620                                         }
 621                                         return nil
 622                                 })
 623                         }
 624                         throttleCPU.Wait()
 625
 626                         var onehotChunk [][]int8
 627                         var onehotXref []onehotXref
 628
 629                         var annotationsFilename string
 630                         if *onlyPCA {
 631                                 annotationsFilename = "/dev/null"
 632                         } else {
 633                                 annotationsFilename = fmt.Sprintf("%s/matrix.%04d.annotations.csv", *outputDir, infileIdx)
 634                                 log.Infof("%04d: writing %s", infileIdx, annotationsFilename)
 635                         }
 636                         annof, err := os.Create(annotationsFilename)
 637                         if err != nil {
 638                                 return err
 639                         }
 640                         annow := bufio.NewWriterSize(annof, 1<<20)
 641                         outcol := 0
 642                         for tag := tagstart; tag < tagend; tag++ {
 643                                 rt := reftile[tag]
 644                                 if rt == nil && mask != nil {
 645                                         // With no ref tile, we don't
 646                                         // have coordinates to say
 647                                         // this is in the desired
 648                                         // regions -- so it's not.
 649                                         // TODO: handle ref spanning
 650                                         // tile case.
 651                                         continue
 652                                 }
 653                                 if rt != nil && rt.excluded {
 654                                         // TODO: don't skip yet --
 655                                         // first check for spanning
 656                                         // tile variants that
 657                                         // intersect non-excluded ref
 658                                         // tiles.
 659                                         continue
 660                                 }
 661                                 if cmd.filter.MaxTag >= 0 && tag > tagID(cmd.filter.MaxTag) {
 662                                         break
 663                                 }
 664                                 remap := variantRemap[tag-tagstart]
 665                                 if remap == nil {
 666                                         // was not assigned above,
 667                                         // because minCoverage
 668                                         outcol++
 669                                         continue
 670                                 }
 671                                 maxv := tileVariantID(0)
 672                                 for _, v := range remap {
 673                                         if maxv < v {
 674                                                 maxv = v
 675                                         }
 676                                 }
 677                                 if *onehotChunked || *onehotSingle || *onlyPCA {
 678                                         onehot, xrefs := cmd.tv2homhet(cgs, maxv, remap, tag, tagstart, seq)
 679                                         if tag == cmd.debugTag {
 680                                                 log.WithFields(logrus.Fields{
 681                                                         "onehot": onehot,
 682                                                         "xrefs":  xrefs,
 683                                                 }).Info("tv2homhet()")
 684                                         }
 685                                         onehotChunk = append(onehotChunk, onehot...)
 686                                         onehotXref = append(onehotXref, xrefs...)
 687                                 }
 688                                 if *onlyPCA {
 689                                         outcol++
 690                                         continue
 691                                 }
 692                                 if rt == nil {
 693                                         // Reference does not use any
 694                                         // variant of this tile
 695                                         //
 696                                         // TODO: diff against the
 697                                         // relevant portion of the
 698                                         // ref's spanning tile
 699                                         outcol++
 700                                         continue
 701                                 }
 702                                 fmt.Fprintf(annow, "%d,%d,%d,=,%s,%d,,,\n", tag, outcol, rt.variant, rt.seqname, rt.pos)
 703                                 variants := seq[tag]
 704                                 reftilestr := strings.ToUpper(string(rt.tiledata))
 705
 706                                 done := make([]bool, maxv+1)
 707                                 variantDiffs := make([][]hgvs.Variant, maxv+1)
 708                                 for v, tv := range variants {
 709                                         v := remap[v]
 710                                         if v == 0 || v == rt.variant || done[v] {
 711                                                 continue
 712                                         } else {
 713                                                 done[v] = true
 714                                         }
 715                                         if len(tv.Sequence) < taglen {
 716                                                 continue
 717                                         }
 718                                         // if reftilestr doesn't end
 719                                         // in the same tag as tv,
 720                                         // extend reftilestr with
 721                                         // following ref tiles until
 722                                         // it does (up to an arbitrary
 723                                         // sanity-check limit)
 724                                         reftilestr := reftilestr
 725                                         endtagstr := strings.ToUpper(string(tv.Sequence[len(tv.Sequence)-taglen:]))
 726                                         for i, rt := 0, rt; i < annotationMaxTileSpan && !strings.HasSuffix(reftilestr, endtagstr) && rt.nexttag >= 0; i++ {
 727                                                 rt = reftile[rt.nexttag]
 728                                                 if rt == nil {
 729                                                         break
 730                                                 }
 731                                                 reftilestr += strings.ToUpper(string(rt.tiledata[taglen:]))
 732                                         }
 733                                         if mask != nil && !mask.Check(strings.TrimPrefix(rt.seqname, "chr"), rt.pos, rt.pos+len(reftilestr)) {
 734                                                 continue
 735                                         }
 736                                         if !strings.HasSuffix(reftilestr, endtagstr) {
 737                                                 fmt.Fprintf(annow, "%d,%d,%d,,%s,%d,,,\n", tag, outcol, v, rt.seqname, rt.pos)
 738                                                 continue
 739                                         }
 740                                         if lendiff := len(reftilestr) - len(tv.Sequence); lendiff < -1000 || lendiff > 1000 {
 741                                                 fmt.Fprintf(annow, "%d,%d,%d,,%s,%d,,,\n", tag, outcol, v, rt.seqname, rt.pos)
 742                                                 continue
 743                                         }
 744                                         diffs, _ := hgvs.Diff(reftilestr, strings.ToUpper(string(tv.Sequence)), 0)
 745                                         for i := range diffs {
 746                                                 diffs[i].Position += rt.pos
 747                                         }
 748                                         for _, diff := range diffs {
 749                                                 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)
 750                                         }
 751                                         if *hgvsChunked {
 752                                                 variantDiffs[v] = diffs
 753                                         }
 754                                 }
 755                                 if *hgvsChunked {
 756                                         // We can now determine, for each HGVS
 757                                         // variant (diff) in this reftile
 758                                         // region, whether a given genome
 759                                         // phase/allele (1) has the variant, (0) has
 760                                         // =ref or a different variant in that
 761                                         // position, or (-1) is lacking
 762                                         // coverage / couldn't be diffed.
 763                                         hgvsCol := hgvsColSet{}
 764                                         for _, diffs := range variantDiffs {
 765                                                 for _, diff := range diffs {
 766                                                         if _, ok := hgvsCol[diff]; ok {
 767                                                                 continue
 768                                                         }
 769                                                         hgvsCol[diff] = [2][]int8{
 770                                                                 make([]int8, len(cmd.cgnames)),
 771                                                                 make([]int8, len(cmd.cgnames)),
 772                                                         }
 773                                                 }
 774                                         }
 775                                         for row, name := range cmd.cgnames {
 776                                                 variants := cgs[name].Variants[(tag-tagstart)*2:]
 777                                                 for ph := 0; ph < 2; ph++ {
 778                                                         v := variants[ph]
 779                                                         if int(v) >= len(remap) {
 780                                                                 v = 0
 781                                                         } else {
 782                                                                 v = remap[v]
 783                                                         }
 784                                                         if v == rt.variant {
 785                                                                 // hgvsCol[*][ph][row] is already 0
 786                                                         } else if len(variantDiffs[v]) == 0 {
 787                                                                 // lacking coverage / couldn't be diffed
 788                                                                 for _, col := range hgvsCol {
 789                                                                         col[ph][row] = -1
 790                                                                 }
 791                                                         } else {
 792                                                                 for _, diff := range variantDiffs[v] {
 793                                                                         hgvsCol[diff][ph][row] = 1
 794                                                                 }
 795                                                         }
 796                                                 }
 797                                         }
 798                                         for diff, colpair := range hgvsCol {
 799                                                 allele2homhet(colpair)
 800                                                 if !cmd.filterHGVScolpair(colpair) {
 801                                                         delete(hgvsCol, diff)
 802                                                 }
 803                                         }
 804                                         if len(hgvsCol) > 0 {
 805                                                 encodeHGVSTodo[rt.seqname] <- hgvsCol
 806                                         }
 807                                 }
 808                                 outcol++
 809                         }
 810                         err = annow.Flush()
 811                         if err != nil {
 812                                 return err
 813                         }
 814                         err = annof.Close()
 815                         if err != nil {
 816                                 return err
 817                         }
 818
 819                         if *onehotChunked {
 820                                 // transpose onehotChunk[col][row] to numpy[row*ncols+col]
 821                                 rows := len(cmd.cgnames)
 822                                 cols := len(onehotChunk)
 823                                 log.Infof("%04d: preparing onehot numpy (rows=%d, cols=%d, mem=%d)", infileIdx, rows, cols, rows*cols)
 824                                 throttleNumpyMem.Acquire()
 825                                 out := onehotcols2int8(onehotChunk)
 826                                 fnm := fmt.Sprintf("%s/onehot.%04d.npy", *outputDir, infileIdx)
 827                                 err = writeNumpyInt8(fnm, out, rows, cols)
 828                                 if err != nil {
 829                                         return err
 830                                 }
 831                                 fnm = fmt.Sprintf("%s/onehot-columns.%04d.npy", *outputDir, infileIdx)
 832                                 err = writeNumpyInt32(fnm, onehotXref2int32(onehotXref), 4, len(onehotXref))
 833                                 if err != nil {
 834                                         return err
 835                                 }
 836                                 debug.FreeOSMemory()
 837                                 throttleNumpyMem.Release()
 838                         }
 839                         if *onehotSingle || *onlyPCA {
 840                                 onehotIndirect[infileIdx] = onehotChunk2Indirect(onehotChunk)
 841                                 onehotChunkSize[infileIdx] = uint32(len(onehotChunk))
 842                                 onehotXrefs[infileIdx] = onehotXref
 843                                 n := len(onehotIndirect[infileIdx][0])
 844                                 log.Infof("%04d: keeping onehot coordinates in memory (n=%d, mem=%d)", infileIdx, n, n*8*2)
 845                         }
 846                         if !(*onehotSingle || *onehotChunked || *onlyPCA) || *mergeOutput || *hgvsSingle {
 847                                 log.Infof("%04d: preparing numpy (rows=%d, cols=%d)", infileIdx, len(cmd.cgnames), 2*outcol)
 848                                 throttleNumpyMem.Acquire()
 849                                 rows := len(cmd.cgnames)
 850                                 cols := 2 * outcol
 851                                 out := make([]int16, rows*cols)
 852                                 for row, name := range cmd.cgnames {
 853                                         outidx := row * cols
 854                                         for col, v := range cgs[name].Variants {
 855                                                 tag := tagstart + tagID(col/2)
 856                                                 if cmd.filter.MaxTag >= 0 && tag > tagID(cmd.filter.MaxTag) {
 857                                                         break
 858                                                 }
 859                                                 if rt := reftile[tag]; rt == nil || rt.excluded {
 860                                                         continue
 861                                                 }
 862                                                 if v == 0 {
 863                                                         out[outidx] = 0 // tag not found / spanning tile
 864                                                 } else if variants, ok := seq[tag]; ok && int(v) < len(variants) && len(variants[v].Sequence) > 0 {
 865                                                         out[outidx] = int16(variantRemap[tag-tagstart][v])
 866                                                 } else {
 867                                                         out[outidx] = -1 // low quality tile variant
 868                                                 }
 869                                                 if tag == cmd.debugTag {
 870                                                         log.Printf("tag %d row %d col %d outidx %d v %d out %d", tag, row, col, outidx, v, out[outidx])
 871                                                 }
 872                                                 outidx++
 873                                         }
 874                                 }
 875                                 seq = nil
 876                                 cgs = nil
 877                                 debug.FreeOSMemory()
 878                                 throttleNumpyMem.Release()
 879                                 if *mergeOutput || *hgvsSingle {
 880                                         log.Infof("%04d: matrix fragment %d rows x %d cols", infileIdx, rows, cols)
 881                                         toMerge[infileIdx] = out
 882                                 }
 883                                 if !*mergeOutput && !*onehotChunked && !*onehotSingle {
 884                                         fnm := fmt.Sprintf("%s/matrix.%04d.npy", *outputDir, infileIdx)
 885                                         err = writeNumpyInt16(fnm, out, rows, cols)
 886                                         if err != nil {
 887                                                 return err
 888                                         }
 889                                 }
 890                         }
 891                         debug.FreeOSMemory()
 892                         log.Infof("%s: done (%d/%d)", infile, int(atomic.AddInt64(&done, 1)), len(infiles))
 893                         return nil
 894                 })
 895         }
 896         if err = throttleMem.Wait(); err != nil {
 897                 return err
 898         }
 899
 900         if *hgvsChunked {
 901                 log.Info("flushing hgvsCols temp files")
 902                 for seqname := range refseq {
 903                         close(encodeHGVSTodo[seqname])
 904                 }
 905                 err = encodeHGVS.Wait()
 906                 if err != nil {
 907                         return err
 908                 }
 909                 for seqname := range refseq {
 910                         log.Infof("%s: reading hgvsCols from temp file", seqname)
 911                         f := tmpHGVSCols[seqname]
 912                         _, err = f.Seek(0, io.SeekStart)
 913                         if err != nil {
 914                                 return err
 915                         }
 916                         var hgvsCols hgvsColSet
 917                         dec := gob.NewDecoder(bufio.NewReaderSize(f, 1<<24))
 918                         for err == nil {
 919                                 err = dec.Decode(&hgvsCols)
 920                         }
 921                         if err != io.EOF {
 922                                 return err
 923                         }
 924                         log.Infof("%s: sorting %d hgvs variants", seqname, len(hgvsCols))
 925                         variants := make([]hgvs.Variant, 0, len(hgvsCols))
 926                         for v := range hgvsCols {
 927                                 variants = append(variants, v)
 928                         }
 929                         sort.Slice(variants, func(i, j int) bool {
 930                                 vi, vj := &variants[i], &variants[j]
 931                                 if vi.Position != vj.Position {
 932                                         return vi.Position < vj.Position
 933                                 } else if vi.Ref != vj.Ref {
 934                                         return vi.Ref < vj.Ref
 935                                 } else {
 936                                         return vi.New < vj.New
 937                                 }
 938                         })
 939                         rows := len(cmd.cgnames)
 940                         cols := len(variants) * 2
 941                         log.Infof("%s: building hgvs matrix (rows=%d, cols=%d, mem=%d)", seqname, rows, cols, rows*cols)
 942                         out := make([]int8, rows*cols)
 943                         for varIdx, variant := range variants {
 944                                 hgvsCols := hgvsCols[variant]
 945                                 for row := range cmd.cgnames {
 946                                         for ph := 0; ph < 2; ph++ {
 947                                                 out[row*cols+varIdx+ph] = hgvsCols[ph][row]
 948                                         }
 949                                 }
 950                         }
 951                         err = writeNumpyInt8(fmt.Sprintf("%s/hgvs.%s.npy", *outputDir, seqname), out, rows, cols)
 952                         if err != nil {
 953                                 return err
 954                         }
 955                         out = nil
 956
 957                         fnm := fmt.Sprintf("%s/hgvs.%s.annotations.csv", *outputDir, seqname)
 958                         log.Infof("%s: writing hgvs column labels to %s", seqname, fnm)
 959                         var hgvsLabels bytes.Buffer
 960                         for varIdx, variant := range variants {
 961                                 fmt.Fprintf(&hgvsLabels, "%d,%s:g.%s\n", varIdx, seqname, variant.String())
 962                         }
 963                         err = ioutil.WriteFile(fnm, hgvsLabels.Bytes(), 0666)
 964                         if err != nil {
 965                                 return err
 966                         }
 967                 }
 968         }
 969
 970         if *mergeOutput || *hgvsSingle {
 971                 var annow *bufio.Writer
 972                 var annof *os.File
 973                 if *mergeOutput {
 974                         annoFilename := fmt.Sprintf("%s/matrix.annotations.csv", *outputDir)
 975                         annof, err = os.Create(annoFilename)
 976                         if err != nil {
 977                                 return err
 978                         }
 979                         annow = bufio.NewWriterSize(annof, 1<<20)
 980                 }
 981
 982                 rows := len(cmd.cgnames)
 983                 cols := 0
 984                 for _, chunk := range toMerge {
 985                         cols += len(chunk) / rows
 986                 }
 987                 log.Infof("merging output matrix (rows=%d, cols=%d, mem=%d) and annotations", rows, cols, rows*cols*2)
 988                 var out []int16
 989                 if *mergeOutput {
 990                         out = make([]int16, rows*cols)
 991                 }
 992                 hgvsCols := map[string][2][]int16{} // hgvs -> [[g0,g1,g2,...], [g0,g1,g2,...]] (slice of genomes for each phase)
 993                 startcol := 0
 994                 for outIdx, chunk := range toMerge {
 995                         chunkcols := len(chunk) / rows
 996                         if *mergeOutput {
 997                                 for row := 0; row < rows; row++ {
 998                                         copy(out[row*cols+startcol:], chunk[row*chunkcols:(row+1)*chunkcols])
 999                                 }
1000                         }
1001                         toMerge[outIdx] = nil
1002
1003                         annotationsFilename := fmt.Sprintf("%s/matrix.%04d.annotations.csv", *outputDir, outIdx)
1004                         log.Infof("reading %s", annotationsFilename)
1005                         buf, err := os.ReadFile(annotationsFilename)
1006                         if err != nil {
1007                                 return err
1008                         }
1009                         if *mergeOutput {
1010                                 err = os.Remove(annotationsFilename)
1011                                 if err != nil {
1012                                         return err
1013                                 }
1014                         }
1015                         for _, line := range bytes.Split(buf, []byte{'\n'}) {
1016                                 if len(line) == 0 {
1017                                         continue
1018                                 }
1019                                 fields := bytes.SplitN(line, []byte{','}, 9)
1020                                 tag, _ := strconv.Atoi(string(fields[0]))
1021                                 incol, _ := strconv.Atoi(string(fields[1]))
1022                                 tileVariant, _ := strconv.Atoi(string(fields[2]))
1023                                 hgvsID := string(fields[3])
1024                                 seqname := string(fields[4])
1025                                 pos, _ := strconv.Atoi(string(fields[5]))
1026                                 refseq := fields[6]
1027                                 if hgvsID == "" {
1028                                         // Null entry for un-diffable
1029                                         // tile variant
1030                                         continue
1031                                 }
1032                                 if hgvsID == "=" {
1033                                         // Null entry for ref tile
1034                                         continue
1035                                 }
1036                                 if mask != nil && !mask.Check(strings.TrimPrefix(seqname, "chr"), pos, pos+len(refseq)) {
1037                                         // The tile intersects one of
1038                                         // the selected regions, but
1039                                         // this particular HGVS
1040                                         // variant does not.
1041                                         continue
1042                                 }
1043                                 hgvsColPair := hgvsCols[hgvsID]
1044                                 if hgvsColPair[0] == nil {
1045                                         // values in new columns start
1046                                         // out as -1 ("no data yet")
1047                                         // or 0 ("=ref") here, may
1048                                         // change to 1 ("hgvs variant
1049                                         // present") below, either on
1050                                         // this line or a future line.
1051                                         hgvsColPair = [2][]int16{make([]int16, len(cmd.cgnames)), make([]int16, len(cmd.cgnames))}
1052                                         rt, ok := reftile[tagID(tag)]
1053                                         if !ok {
1054                                                 err = fmt.Errorf("bug: seeing annotations for tag %d, but it has no reftile entry", tag)
1055                                                 return err
1056                                         }
1057                                         for ph := 0; ph < 2; ph++ {
1058                                                 for row := 0; row < rows; row++ {
1059                                                         v := chunk[row*chunkcols+incol*2+ph]
1060                                                         if tileVariantID(v) == rt.variant {
1061                                                                 hgvsColPair[ph][row] = 0
1062                                                         } else {
1063                                                                 hgvsColPair[ph][row] = -1
1064                                                         }
1065                                                 }
1066                                         }
1067                                         hgvsCols[hgvsID] = hgvsColPair
1068                                         if annow != nil {
1069                                                 hgvsref := hgvs.Variant{
1070                                                         Position: pos,
1071                                                         Ref:      string(refseq),
1072                                                         New:      string(refseq),
1073                                                 }
1074                                                 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])
1075                                         }
1076                                 }
1077                                 if annow != nil {
1078                                         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])
1079                                 }
1080                                 for ph := 0; ph < 2; ph++ {
1081                                         for row := 0; row < rows; row++ {
1082                                                 v := chunk[row*chunkcols+incol*2+ph]
1083                                                 if int(v) == tileVariant {
1084                                                         hgvsColPair[ph][row] = 1
1085                                                 }
1086                                         }
1087                                 }
1088                         }
1089
1090                         startcol += chunkcols
1091                 }
1092                 if *mergeOutput {
1093                         err = annow.Flush()
1094                         if err != nil {
1095                                 return err
1096                         }
1097                         err = annof.Close()
1098                         if err != nil {
1099                                 return err
1100                         }
1101                         err = writeNumpyInt16(fmt.Sprintf("%s/matrix.npy", *outputDir), out, rows, cols)
1102                         if err != nil {
1103                                 return err
1104                         }
1105                 }
1106                 out = nil
1107
1108                 if *hgvsSingle {
1109                         cols = len(hgvsCols) * 2
1110                         log.Printf("building hgvs-based matrix: %d rows x %d cols", rows, cols)
1111                         out = make([]int16, rows*cols)
1112                         hgvsIDs := make([]string, 0, cols/2)
1113                         for hgvsID := range hgvsCols {
1114                                 hgvsIDs = append(hgvsIDs, hgvsID)
1115                         }
1116                         sort.Strings(hgvsIDs)
1117                         var hgvsLabels bytes.Buffer
1118                         for idx, hgvsID := range hgvsIDs {
1119                                 fmt.Fprintf(&hgvsLabels, "%d,%s\n", idx, hgvsID)
1120                                 for ph := 0; ph < 2; ph++ {
1121                                         hgvscol := hgvsCols[hgvsID][ph]
1122                                         for row, val := range hgvscol {
1123                                                 out[row*cols+idx*2+ph] = val
1124                                         }
1125                                 }
1126                         }
1127                         err = writeNumpyInt16(fmt.Sprintf("%s/hgvs.npy", *outputDir), out, rows, cols)
1128                         if err != nil {
1129                                 return err
1130                         }
1131
1132                         fnm := fmt.Sprintf("%s/hgvs.annotations.csv", *outputDir)
1133                         log.Printf("writing hgvs labels: %s", fnm)
1134                         err = ioutil.WriteFile(fnm, hgvsLabels.Bytes(), 0777)
1135                         if err != nil {
1136                                 return err
1137                         }
1138                 }
1139         }
1140         if *onehotSingle || *onlyPCA {
1141                 nzCount := 0
1142                 for _, part := range onehotIndirect {
1143                         nzCount += len(part[0])
1144                 }
1145                 onehot := make([]uint32, nzCount*2) // [r,r,r,...,c,c,c,...]
1146                 var xrefs []onehotXref
1147                 chunkOffset := uint32(0)
1148                 outcol := 0
1149                 for i, part := range onehotIndirect {
1150                         for i := range part[1] {
1151                                 part[1][i] += chunkOffset
1152                         }
1153                         copy(onehot[outcol:], part[0])
1154                         copy(onehot[outcol+nzCount:], part[1])
1155                         xrefs = append(xrefs, onehotXrefs[i]...)
1156
1157                         outcol += len(part[0])
1158                         chunkOffset += onehotChunkSize[i]
1159
1160                         part[0] = nil
1161                         part[1] = nil
1162                         onehotXrefs[i] = nil
1163                         debug.FreeOSMemory()
1164                 }
1165                 if *onehotSingle {
1166                         fnm := fmt.Sprintf("%s/onehot.npy", *outputDir)
1167                         err = writeNumpyUint32(fnm, onehot, 2, nzCount)
1168                         if err != nil {
1169                                 return err
1170                         }
1171                         fnm = fmt.Sprintf("%s/onehot-columns.npy", *outputDir)
1172                         err = writeNumpyInt32(fnm, onehotXref2int32(xrefs), 5, len(xrefs))
1173                         if err != nil {
1174                                 return err
1175                         }
1176                 }
1177                 if *onlyPCA {
1178                         cols := 0
1179                         for _, c := range onehot[nzCount:] {
1180                                 if int(c) >= cols {
1181                                         cols = int(c) + 1
1182                                 }
1183                         }
1184                         if cols == 0 {
1185                                 return fmt.Errorf("cannot do PCA: one-hot matrix is empty")
1186                         }
1187                         log.Printf("have %d one-hot cols", cols)
1188                         stride := 1
1189                         for *maxPCATiles > 0 && cols > *maxPCATiles*2 {
1190                                 cols = (cols + 1) / 2
1191                                 stride = stride * 2
1192                         }
1193                         if cols%2 == 1 {
1194                                 // we work with pairs of columns
1195                                 cols++
1196                         }
1197                         log.Printf("creating full matrix (%d rows) and training matrix (%d rows) with %d cols, stride %d", len(cmd.cgnames), cmd.trainingSetSize, cols, stride)
1198                         mtxFull := mat.NewDense(len(cmd.cgnames), cols, nil)
1199                         mtxTrain := mat.NewDense(cmd.trainingSetSize, cols, nil)
1200                         for i, c := range onehot[nzCount:] {
1201                                 if int(c/2)%stride == 0 {
1202                                         outcol := int(c/2)/stride*2 + int(c)%2
1203                                         mtxFull.Set(int(onehot[i]), outcol, 1)
1204                                         if trainRow := cmd.trainingSet[int(onehot[i])]; trainRow >= 0 {
1205                                                 mtxTrain.Set(trainRow, outcol, 1)
1206                                         }
1207                                 }
1208                         }
1209                         log.Print("fitting")
1210                         transformer := nlp.NewPCA(*pcaComponents)
1211                         transformer.Fit(mtxTrain.T())
1212                         log.Printf("transforming")
1213                         pca, err := transformer.Transform(mtxFull.T())
1214                         if err != nil {
1215                                 return err
1216                         }
1217                         pca = pca.T()
1218                         outrows, outcols := pca.Dims()
1219                         log.Printf("copying result to numpy output array: %d rows, %d cols", outrows, outcols)
1220                         out := make([]float64, outrows*outcols)
1221                         for i := 0; i < outrows; i++ {
1222                                 for j := 0; j < outcols; j++ {
1223                                         out[i*outcols+j] = pca.At(i, j)
1224                                 }
1225                         }
1226                         fnm := fmt.Sprintf("%s/pca.npy", *outputDir)
1227                         log.Printf("writing numpy: %s", fnm)
1228                         output, err := os.OpenFile(fnm, os.O_CREATE|os.O_TRUNC|os.O_WRONLY, 0777)
1229                         if err != nil {
1230                                 return err
1231                         }
1232                         npw, err := gonpy.NewWriter(nopCloser{output})
1233                         if err != nil {
1234                                 return fmt.Errorf("gonpy.NewWriter: %w", err)
1235                         }
1236                         npw.Shape = []int{outrows, outcols}
1237                         err = npw.WriteFloat64(out)
1238                         if err != nil {
1239                                 return fmt.Errorf("WriteFloat64: %w", err)
1240                         }
1241                         err = output.Close()
1242                         if err != nil {
1243                                 return err
1244                         }
1245                         log.Print("done")
1246
1247                         samplesOutFilename := *outputDir + "/samples.csv"
1248                         log.Infof("writing sample metadata to %s", samplesOutFilename)
1249                         var f *os.File
1250                         f, err = os.Create(samplesOutFilename)
1251                         if err != nil {
1252                                 return err
1253                         }
1254                         defer f.Close()
1255                         for i, si := range cmd.samples {
1256                                 var cc, tv string
1257                                 if si.isCase {
1258                                         cc = "1"
1259                                 } else if si.isControl {
1260                                         cc = "0"
1261                                 }
1262                                 if si.isTraining {
1263                                         tv = "1"
1264                                 } else {
1265                                         tv = "0"
1266                                 }
1267                                 var pcavals string
1268                                 for c := 0; c < outcols; c++ {
1269                                         pcavals += fmt.Sprintf(",%f", pca.At(i, c))
1270                                 }
1271                                 _, err = fmt.Fprintf(f, "%d,%s,%s,%s%s\n", i, si.id, cc, tv, pcavals)
1272                                 if err != nil {
1273                                         err = fmt.Errorf("write %s: %w", samplesOutFilename, err)
1274                                         return err
1275                                 }
1276                         }
1277                         err = f.Close()
1278                         if err != nil {
1279                                 err = fmt.Errorf("close %s: %w", samplesOutFilename, err)
1280                                 return err
1281                         }
1282                         log.Print("done")
1283                 }
1284         }
1285         if !*mergeOutput && !*onehotChunked && !*onehotSingle && !*onlyPCA {
1286                 tagoffsetFilename := *outputDir + "/chunk-tag-offset.csv"
1287                 log.Infof("writing tag offsets to %s", tagoffsetFilename)
1288                 var f *os.File
1289                 f, err = os.Create(tagoffsetFilename)
1290                 if err != nil {
1291                         return err
1292                 }
1293                 defer f.Close()
1294                 for idx, offset := range chunkStartTag {
1295                         _, err = fmt.Fprintf(f, "%q,%d\n", fmt.Sprintf("matrix.%04d.npy", idx), offset)
1296                         if err != nil {
1297                                 err = fmt.Errorf("write %s: %w", tagoffsetFilename, err)
1298                                 return err
1299                         }
1300                 }
1301                 err = f.Close()
1302                 if err != nil {
1303                         err = fmt.Errorf("close %s: %w", tagoffsetFilename, err)
1304                         return err
1305                 }
1306         }
1307
1308         return nil
1309 }
1310
1311 type sampleInfo struct {
1312         id            string
1313         isCase        bool
1314         isControl     bool
1315         isTraining    bool
1316         isValidation  bool
1317         pcaComponents []float64
1318 }
1319
1320 // Read samples.csv file with case/control and training/validation
1321 // flags.
1322 func (cmd *sliceNumpy) loadSampleInfo(samplesFilename string) ([]sampleInfo, error) {
1323         var si []sampleInfo
1324         f, err := open(samplesFilename)
1325         if err != nil {
1326                 return nil, err
1327         }
1328         buf, err := io.ReadAll(f)
1329         f.Close()
1330         if err != nil {
1331                 return nil, err
1332         }
1333         lineNum := 0
1334         for _, csv := range bytes.Split(buf, []byte{'\n'}) {
1335                 lineNum++
1336                 if len(csv) == 0 {
1337                         continue
1338                 }
1339                 split := strings.Split(string(csv), ",")
1340                 if len(split) < 4 {
1341                         return nil, fmt.Errorf("%d fields < 4 in %s line %d: %q", len(split), samplesFilename, lineNum, csv)
1342                 }
1343                 if split[0] == "Index" && split[1] == "SampleID" && split[2] == "CaseControl" && split[3] == "TrainingValidation" {
1344                         continue
1345                 }
1346                 idx, err := strconv.Atoi(split[0])
1347                 if err != nil {
1348                         if lineNum == 1 {
1349                                 return nil, fmt.Errorf("header does not look right: %q", csv)
1350                         }
1351                         return nil, fmt.Errorf("%s line %d: index: %s", samplesFilename, lineNum, err)
1352                 }
1353                 if idx != len(si) {
1354                         return nil, fmt.Errorf("%s line %d: index %d out of order", samplesFilename, lineNum, idx)
1355                 }
1356                 var pcaComponents []float64
1357                 if len(split) > 4 {
1358                         for _, s := range split[4:] {
1359                                 f, err := strconv.ParseFloat(s, 64)
1360                                 if err != nil {
1361                                         return nil, fmt.Errorf("%s line %d: cannot parse float %q: %s", samplesFilename, lineNum, s, err)
1362                                 }
1363                                 pcaComponents = append(pcaComponents, f)
1364                         }
1365                 }
1366                 si = append(si, sampleInfo{
1367                         id:            split[1],
1368                         isCase:        split[2] == "1",
1369                         isControl:     split[2] == "0",
1370                         isTraining:    split[3] == "1",
1371                         isValidation:  split[3] == "0",
1372                         pcaComponents: pcaComponents,
1373                 })
1374         }
1375         return si, nil
1376 }
1377
1378 func (cmd *sliceNumpy) filterHGVScolpair(colpair [2][]int8) bool {
1379         if cmd.chi2PValue >= 1 {
1380                 return true
1381         }
1382         col0 := make([]bool, 0, len(cmd.chi2Cases))
1383         col1 := make([]bool, 0, len(cmd.chi2Cases))
1384         cases := make([]bool, 0, len(cmd.chi2Cases))
1385         for i, c := range cmd.chi2Cases {
1386                 if colpair[0][i] < 0 {
1387                         continue
1388                 }
1389                 col0 = append(col0, colpair[0][i] != 0)
1390                 col1 = append(col1, colpair[1][i] != 0)
1391                 cases = append(cases, c)
1392         }
1393         return len(cases) >= cmd.minCoverage &&
1394                 (pvalue(col0, cases) <= cmd.chi2PValue || pvalue(col1, cases) <= cmd.chi2PValue)
1395 }
1396
1397 func writeNumpyUint32(fnm string, out []uint32, rows, cols int) error {
1398         output, err := os.Create(fnm)
1399         if err != nil {
1400                 return err
1401         }
1402         defer output.Close()
1403         bufw := bufio.NewWriterSize(output, 1<<26)
1404         npw, err := gonpy.NewWriter(nopCloser{bufw})
1405         if err != nil {
1406                 return err
1407         }
1408         log.WithFields(log.Fields{
1409                 "filename": fnm,
1410                 "rows":     rows,
1411                 "cols":     cols,
1412                 "bytes":    rows * cols * 4,
1413         }).Infof("writing numpy: %s", fnm)
1414         npw.Shape = []int{rows, cols}
1415         npw.WriteUint32(out)
1416         err = bufw.Flush()
1417         if err != nil {
1418                 return err
1419         }
1420         return output.Close()
1421 }
1422
1423 func writeNumpyInt32(fnm string, out []int32, rows, cols int) error {
1424         output, err := os.Create(fnm)
1425         if err != nil {
1426                 return err
1427         }
1428         defer output.Close()
1429         bufw := bufio.NewWriterSize(output, 1<<26)
1430         npw, err := gonpy.NewWriter(nopCloser{bufw})
1431         if err != nil {
1432                 return err
1433         }
1434         log.WithFields(log.Fields{
1435                 "filename": fnm,
1436                 "rows":     rows,
1437                 "cols":     cols,
1438                 "bytes":    rows * cols * 4,
1439         }).Infof("writing numpy: %s", fnm)
1440         npw.Shape = []int{rows, cols}
1441         npw.WriteInt32(out)
1442         err = bufw.Flush()
1443         if err != nil {
1444                 return err
1445         }
1446         return output.Close()
1447 }
1448
1449 func writeNumpyInt16(fnm string, out []int16, rows, cols int) error {
1450         output, err := os.Create(fnm)
1451         if err != nil {
1452                 return err
1453         }
1454         defer output.Close()
1455         bufw := bufio.NewWriterSize(output, 1<<26)
1456         npw, err := gonpy.NewWriter(nopCloser{bufw})
1457         if err != nil {
1458                 return err
1459         }
1460         log.WithFields(log.Fields{
1461                 "filename": fnm,
1462                 "rows":     rows,
1463                 "cols":     cols,
1464                 "bytes":    rows * cols * 2,
1465         }).Infof("writing numpy: %s", fnm)
1466         npw.Shape = []int{rows, cols}
1467         npw.WriteInt16(out)
1468         err = bufw.Flush()
1469         if err != nil {
1470                 return err
1471         }
1472         return output.Close()
1473 }
1474
1475 func writeNumpyInt8(fnm string, out []int8, rows, cols int) error {
1476         output, err := os.Create(fnm)
1477         if err != nil {
1478                 return err
1479         }
1480         defer output.Close()
1481         bufw := bufio.NewWriterSize(output, 1<<26)
1482         npw, err := gonpy.NewWriter(nopCloser{bufw})
1483         if err != nil {
1484                 return err
1485         }
1486         log.WithFields(log.Fields{
1487                 "filename": fnm,
1488                 "rows":     rows,
1489                 "cols":     cols,
1490                 "bytes":    rows * cols,
1491         }).Infof("writing numpy: %s", fnm)
1492         npw.Shape = []int{rows, cols}
1493         npw.WriteInt8(out)
1494         err = bufw.Flush()
1495         if err != nil {
1496                 return err
1497         }
1498         return output.Close()
1499 }
1500
1501 func allele2homhet(colpair [2][]int8) {
1502         a, b := colpair[0], colpair[1]
1503         for i, av := range a {
1504                 bv := b[i]
1505                 if av < 0 || bv < 0 {
1506                         // no-call
1507                         a[i], b[i] = -1, -1
1508                 } else if av > 0 && bv > 0 {
1509                         // hom
1510                         a[i], b[i] = 1, 0
1511                 } else if av > 0 || bv > 0 {
1512                         // het
1513                         a[i], b[i] = 0, 1
1514                 } else {
1515                         // ref (or a different variant in same position)
1516                         // (this is a no-op) a[i], b[i] = 0, 0
1517                 }
1518         }
1519 }
1520
1521 type onehotXref struct {
1522         tag     tagID
1523         variant tileVariantID
1524         hom     bool
1525         pvalue  float64
1526 }
1527
1528 const onehotXrefSize = unsafe.Sizeof(onehotXref{})
1529
1530 // Build onehot matrix (m[tileVariantIndex][genome] == 0 or 1) for all
1531 // variants of a single tile/tag#.
1532 //
1533 // Return nil if no tile variant passes Χ² filter.
1534 func (cmd *sliceNumpy) tv2homhet(cgs map[string]CompactGenome, maxv tileVariantID, remap []tileVariantID, tag, chunkstarttag tagID, seq map[tagID][]TileVariant) ([][]int8, []onehotXref) {
1535         if tag == cmd.debugTag {
1536                 tv := make([]tileVariantID, len(cmd.cgnames)*2)
1537                 for i, name := range cmd.cgnames {
1538                         copy(tv[i*2:(i+1)*2], cgs[name].Variants[(tag-chunkstarttag)*2:])
1539                 }
1540                 log.WithFields(logrus.Fields{
1541                         "cgs[i].Variants[tag*2+j]": tv,
1542                         "maxv":                     maxv,
1543                         "remap":                    remap,
1544                         "tag":                      tag,
1545                         "chunkstarttag":            chunkstarttag,
1546                 }).Info("tv2homhet()")
1547         }
1548         if maxv < 1 || (maxv < 2 && !cmd.includeVariant1) {
1549                 // everyone has the most common variant (of the variants we don't drop)
1550                 return nil, nil
1551         }
1552         tagoffset := tag - chunkstarttag
1553         coverage := 0
1554         for _, cg := range cgs {
1555                 alleles := 0
1556                 for _, v := range cg.Variants[tagoffset*2 : tagoffset*2+2] {
1557                         if v > 0 && int(v) < len(seq[tag]) && len(seq[tag][v].Sequence) > 0 {
1558                                 alleles++
1559                         }
1560                 }
1561                 if alleles == 2 {
1562                         coverage++
1563                 }
1564         }
1565         if coverage < cmd.minCoverage {
1566                 return nil, nil
1567         }
1568         // "observed" array for p-value calculation (training set
1569         // only)
1570         obs := make([][]bool, (maxv+1)*2) // 2 slices (hom + het) for each variant#
1571         // one-hot output (all samples)
1572         outcols := make([][]int8, (maxv+1)*2)
1573         for i := range obs {
1574                 obs[i] = make([]bool, cmd.trainingSetSize)
1575                 outcols[i] = make([]int8, len(cmd.cgnames))
1576         }
1577         for cgid, name := range cmd.cgnames {
1578                 tsid := cmd.trainingSet[cgid]
1579                 cgvars := cgs[name].Variants[tagoffset*2:]
1580                 tv0, tv1 := remap[cgvars[0]], remap[cgvars[1]]
1581                 for v := tileVariantID(1); v <= maxv; v++ {
1582                         if tv0 == v && tv1 == v {
1583                                 if tsid >= 0 {
1584                                         obs[v*2][tsid] = true
1585                                 }
1586                                 outcols[v*2][cgid] = 1
1587                         } else if tv0 == v || tv1 == v {
1588                                 if tsid >= 0 {
1589                                         obs[v*2+1][tsid] = true
1590                                 }
1591                                 outcols[v*2+1][cgid] = 1
1592                         }
1593                 }
1594         }
1595         var onehot [][]int8
1596         var xref []onehotXref
1597         for col := 2; col < len(obs); col++ {
1598                 // col 0,1 correspond to tile variant 0, i.e.,
1599                 // no-call; col 2,3 correspond to the most common
1600                 // variant; so we (normally) start at col 4.
1601                 if col < 4 && !cmd.includeVariant1 {
1602                         continue
1603                 }
1604                 var p float64
1605                 if len(cmd.samples[0].pcaComponents) > 0 {
1606                         p = pvalueGLM(cmd.samples, obs[col:col+2])
1607                 } else {
1608                         p = pvalue(obs[col], cmd.chi2Cases)
1609                 }
1610                 if cmd.chi2PValue < 1 && !(p < cmd.chi2PValue) {
1611                         continue
1612                 }
1613                 onehot = append(onehot, outcols[col])
1614                 xref = append(xref, onehotXref{
1615                         tag:     tag,
1616                         variant: tileVariantID(col >> 1),
1617                         hom:     col&1 == 0,
1618                         pvalue:  p,
1619                 })
1620         }
1621         return onehot, xref
1622 }
1623
1624 // convert a []onehotXref with length N to a numpy-style []int32
1625 // matrix with N columns, one row per field of onehotXref struct.
1626 //
1627 // Hom/het row contains hom=0, het=1.
1628 //
1629 // P-value row contains 1000000x actual p-value.
1630 func onehotXref2int32(xrefs []onehotXref) []int32 {
1631         xcols := len(xrefs)
1632         xdata := make([]int32, 5*xcols)
1633         for i, xref := range xrefs {
1634                 xdata[i] = int32(xref.tag)
1635                 xdata[xcols+i] = int32(xref.variant)
1636                 if xref.hom {
1637                         xdata[xcols*2+i] = 1
1638                 }
1639                 xdata[xcols*3+i] = int32(xref.pvalue * 1000000)
1640                 xdata[xcols*4+i] = int32(-math.Log10(xref.pvalue) * 1000000)
1641         }
1642         return xdata
1643 }
1644
1645 // transpose onehot data from in[col][row] to numpy-style
1646 // out[row*cols+col].
1647 func onehotcols2int8(in [][]int8) []int8 {
1648         if len(in) == 0 {
1649                 return nil
1650         }
1651         cols := len(in)
1652         rows := len(in[0])
1653         out := make([]int8, rows*cols)
1654         for row := 0; row < rows; row++ {
1655                 outrow := out[row*cols:]
1656                 for col, incol := range in {
1657                         outrow[col] = incol[row]
1658                 }
1659         }
1660         return out
1661 }
1662
1663 // Return [2][]uint32{rowIndices, colIndices} indicating which
1664 // elements of matrixT[c][r] have non-zero values.
1665 func onehotChunk2Indirect(matrixT [][]int8) [2][]uint32 {
1666         var nz [2][]uint32
1667         for c, col := range matrixT {
1668                 for r, val := range col {
1669                         if val != 0 {
1670                                 nz[0] = append(nz[0], uint32(r))
1671                                 nz[1] = append(nz[1], uint32(c))
1672                         }
1673                 }
1674         }
1675         return nz
1676 }