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