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         onlyPCA := flags.Bool("pca", false, "generate pca matrix")
  86         pcaComponents := flags.Int("pca-components", 4, "number of PCA components")
  87         maxPCATiles := flags.Int("max-pca-tiles", 0, "maximum tiles to use as PCA input (filter, then drop every 2nd colum pair until below max)")
  88         debugTag := flags.Int("debug-tag", -1, "log debugging details about specified tag")
  89         flags.IntVar(&cmd.threads, "threads", 16, "number of memory-hungry assembly threads, and number of VCPUs to request for arvados container")
  90         flags.Float64Var(&cmd.chi2PValue, "chi2-p-value", 1, "do Χ² test and omit columns with p-value above this threshold")
  91         flags.BoolVar(&cmd.includeVariant1, "include-variant-1", false, "include most common variant when building one-hot matrix")
  92         cmd.filter.Flags(flags)
  93         err := flags.Parse(args)
  94         if err == flag.ErrHelp {
  95                 return nil
  96         } else if err != nil {
  97                 return err
  98         }
  99
 100         if *pprof != "" {
 101                 go func() {
 102                         log.Println(http.ListenAndServe(*pprof, nil))
 103                 }()
 104         }
 105
 106         if cmd.chi2PValue != 1 && *samplesFilename == "" {
 107                 return fmt.Errorf("cannot use provided -chi2-p-value=%f because -samples= value is empty", cmd.chi2PValue)
 108         }
 109
 110         cmd.debugTag = tagID(*debugTag)
 111
 112         if !*runlocal {
 113                 runner := arvadosContainerRunner{
 114                         Name:        "lightning slice-numpy",
 115                         Client:      arvados.NewClientFromEnv(),
 116                         ProjectUUID: *projectUUID,
 117                         RAM:         int64(*arvadosRAM),
 118                         VCPUs:       *arvadosVCPUs,
 119                         Priority:    *priority,
 120                         KeepCache:   2,
 121                         APIAccess:   true,
 122                 }
 123                 err = runner.TranslatePaths(inputDir, regionsFilename, samplesFilename)
 124                 if err != nil {
 125                         return err
 126                 }
 127                 runner.Args = []string{"slice-numpy", "-local=true",
 128                         "-pprof=:6060",
 129                         "-input-dir=" + *inputDir,
 130                         "-output-dir=/mnt/output",
 131                         "-threads=" + fmt.Sprintf("%d", cmd.threads),
 132                         "-regions=" + *regionsFilename,
 133                         "-expand-regions=" + fmt.Sprintf("%d", *expandRegions),
 134                         "-merge-output=" + fmt.Sprintf("%v", *mergeOutput),
 135                         "-single-hgvs-matrix=" + fmt.Sprintf("%v", *hgvsSingle),
 136                         "-chunked-hgvs-matrix=" + fmt.Sprintf("%v", *hgvsChunked),
 137                         "-single-onehot=" + fmt.Sprintf("%v", *onehotSingle),
 138                         "-chunked-onehot=" + fmt.Sprintf("%v", *onehotChunked),
 139                         "-samples=" + *samplesFilename,
 140                         "-pca=" + fmt.Sprintf("%v", *onlyPCA),
 141                         "-pca-components=" + fmt.Sprintf("%d", *pcaComponents),
 142                         "-max-pca-tiles=" + fmt.Sprintf("%d", *maxPCATiles),
 143                         "-chi2-p-value=" + fmt.Sprintf("%f", cmd.chi2PValue),
 144                         "-include-variant-1=" + fmt.Sprintf("%v", cmd.includeVariant1),
 145                         "-debug-tag=" + fmt.Sprintf("%d", cmd.debugTag),
 146                 }
 147                 runner.Args = append(runner.Args, cmd.filter.Args()...)
 148                 var output string
 149                 output, err = runner.Run()
 150                 if err != nil {
 151                         return err
 152                 }
 153                 fmt.Fprintln(stdout, output)
 154                 return nil
 155         }
 156
 157         infiles, err := allFiles(*inputDir, matchGobFile)
 158         if err != nil {
 159                 return err
 160         }
 161         if len(infiles) == 0 {
 162                 err = fmt.Errorf("no input files found in %s", *inputDir)
 163                 return err
 164         }
 165         sort.Strings(infiles)
 166
 167         var refseq map[string][]tileLibRef
 168         var reftiledata = make(map[tileLibRef][]byte, 11000000)
 169         in0, err := open(infiles[0])
 170         if err != nil {
 171                 return err
 172         }
 173
 174         matchGenome, err := regexp.Compile(cmd.filter.MatchGenome)
 175         if err != nil {
 176                 err = fmt.Errorf("-match-genome: invalid regexp: %q", cmd.filter.MatchGenome)
 177                 return err
 178         }
 179
 180         if *samplesFilename != "" {
 181                 cmd.samples, err = cmd.loadSampleInfo(*samplesFilename)
 182                 if err != nil {
 183                         return err
 184                 }
 185         }
 186
 187         cmd.cgnames = nil
 188         var tagset [][]byte
 189         err = DecodeLibrary(in0, strings.HasSuffix(infiles[0], ".gz"), func(ent *LibraryEntry) error {
 190                 if len(ent.TagSet) > 0 {
 191                         tagset = ent.TagSet
 192                 }
 193                 for _, cseq := range ent.CompactSequences {
 194                         if cseq.Name == *ref || *ref == "" {
 195                                 refseq = cseq.TileSequences
 196                         }
 197                 }
 198                 for _, cg := range ent.CompactGenomes {
 199                         if matchGenome.MatchString(cg.Name) {
 200                                 cmd.cgnames = append(cmd.cgnames, cg.Name)
 201                         }
 202                 }
 203                 for _, tv := range ent.TileVariants {
 204                         if tv.Ref {
 205                                 reftiledata[tileLibRef{tv.Tag, tv.Variant}] = tv.Sequence
 206                         }
 207                 }
 208                 return nil
 209         })
 210         if err != nil {
 211                 return err
 212         }
 213         in0.Close()
 214         if refseq == nil {
 215                 err = fmt.Errorf("%s: reference sequence not found", infiles[0])
 216                 return err
 217         }
 218         if len(tagset) == 0 {
 219                 err = fmt.Errorf("tagset not found")
 220                 return err
 221         }
 222
 223         taglib := &tagLibrary{}
 224         err = taglib.setTags(tagset)
 225         if err != nil {
 226                 return err
 227         }
 228         taglen := taglib.TagLen()
 229         sort.Strings(cmd.cgnames)
 230
 231         if len(cmd.cgnames) == 0 {
 232                 return fmt.Errorf("fatal: 0 matching samples in library, nothing to do")
 233         }
 234         cmd.trainingSet = make([]int, len(cmd.cgnames))
 235         if *samplesFilename == "" {
 236                 cmd.trainingSetSize = len(cmd.cgnames)
 237                 for i, name := range cmd.cgnames {
 238                         cmd.samples = append(cmd.samples, sampleInfo{
 239                                 id:         trimFilenameForLabel(name),
 240                                 isTraining: true,
 241                         })
 242                         cmd.trainingSet[i] = i
 243                 }
 244         } else if len(cmd.cgnames) != len(cmd.samples) {
 245                 return fmt.Errorf("mismatched sample list: %d samples in library, %d in %s", len(cmd.cgnames), len(cmd.samples), *samplesFilename)
 246         } else {
 247                 cmd.trainingSetSize = 0
 248                 for i, name := range cmd.cgnames {
 249                         if s := trimFilenameForLabel(name); s != cmd.samples[i].id {
 250                                 return fmt.Errorf("mismatched sample list: sample %d is %q in library, %q in %s", i, s, cmd.samples[i].id, *samplesFilename)
 251                         }
 252                         if cmd.samples[i].isTraining {
 253                                 cmd.trainingSet[i] = cmd.trainingSetSize
 254                                 cmd.trainingSetSize++
 255                         } else {
 256                                 cmd.trainingSet[i] = -1
 257                         }
 258                 }
 259         }
 260         if cmd.filter.MinCoverage == 1 {
 261                 // In the generic formula below, floating point
 262                 // arithmetic can effectively push the coverage
 263                 // threshold above 1.0, which is impossible/useless.
 264                 // 1.0 needs to mean exactly 100% coverage.
 265                 cmd.minCoverage = len(cmd.cgnames)
 266         } else {
 267                 cmd.minCoverage = int(math.Ceil(cmd.filter.MinCoverage * float64(len(cmd.cgnames))))
 268         }
 269
 270         log.Info("indexing reference tiles")
 271         type reftileinfo struct {
 272                 variant  tileVariantID
 273                 seqname  string // chr1
 274                 pos      int    // distance from start of chromosome to starttag
 275                 tiledata []byte // acgtggcaa...
 276                 excluded bool   // true if excluded by regions file
 277                 nexttag  tagID  // tagID of following tile (-1 for last tag of chromosome)
 278         }
 279         isdup := map[tagID]bool{}
 280         reftile := map[tagID]*reftileinfo{}
 281         for seqname, cseq := range refseq {
 282                 pos := 0
 283                 lastreftag := tagID(-1)
 284                 for _, libref := range cseq {
 285                         if cmd.filter.MaxTag >= 0 && libref.Tag > tagID(cmd.filter.MaxTag) {
 286                                 continue
 287                         }
 288                         tiledata := reftiledata[libref]
 289                         if len(tiledata) == 0 {
 290                                 err = fmt.Errorf("missing tiledata for tag %d variant %d in %s in ref", libref.Tag, libref.Variant, seqname)
 291                                 return err
 292                         }
 293                         foundthistag := false
 294                         taglib.FindAll(tiledata[:len(tiledata)-1], func(tagid tagID, offset, _ int) {
 295                                 if !foundthistag && tagid == libref.Tag {
 296                                         foundthistag = true
 297                                         return
 298                                 }
 299                                 if dupref, ok := reftile[tagid]; ok {
 300                                         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)
 301                                         delete(reftile, tagid)
 302                                 } else {
 303                                         log.Printf("found tag %d at offset %d inside tile variant %+v on %s @ %d", tagid, offset, libref, seqname, pos+offset+1)
 304                                 }
 305                                 isdup[tagid] = true
 306                         })
 307                         if isdup[libref.Tag] {
 308                                 log.Printf("dropping reference tile %+v from %s @ %d, tag not unique", libref, seqname, pos)
 309                         } else if reftile[libref.Tag] != nil {
 310                                 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)
 311                                 delete(reftile, libref.Tag)
 312                                 log.Printf("dropping reference tile %+v from %s @ %d, tag not unique", libref, seqname, pos)
 313                                 isdup[libref.Tag] = true
 314                         } else {
 315                                 reftile[libref.Tag] = &reftileinfo{
 316                                         seqname:  seqname,
 317                                         variant:  libref.Variant,
 318                                         tiledata: tiledata,
 319                                         pos:      pos,
 320                                         nexttag:  -1,
 321                                 }
 322                                 if lastreftag >= 0 {
 323                                         reftile[lastreftag].nexttag = libref.Tag
 324                                 }
 325                                 lastreftag = libref.Tag
 326                         }
 327                         pos += len(tiledata) - taglen
 328                 }
 329                 log.Printf("... %s done, len %d", seqname, pos+taglen)
 330         }
 331
 332         var mask *mask
 333         if *regionsFilename != "" {
 334                 log.Printf("loading regions from %s", *regionsFilename)
 335                 mask, err = makeMask(*regionsFilename, *expandRegions)
 336                 if err != nil {
 337                         return err
 338                 }
 339                 log.Printf("before applying mask, len(reftile) == %d", len(reftile))
 340                 log.Printf("deleting reftile entries for regions outside %d intervals", mask.Len())
 341                 for _, rt := range reftile {
 342                         if !mask.Check(strings.TrimPrefix(rt.seqname, "chr"), rt.pos, rt.pos+len(rt.tiledata)) {
 343                                 rt.excluded = true
 344                         }
 345                 }
 346                 log.Printf("after applying mask, len(reftile) == %d", len(reftile))
 347         }
 348
 349         type hgvsColSet map[hgvs.Variant][2][]int8
 350         encodeHGVS := throttle{Max: len(refseq)}
 351         encodeHGVSTodo := map[string]chan hgvsColSet{}
 352         tmpHGVSCols := map[string]*os.File{}
 353         if *hgvsChunked {
 354                 for seqname := range refseq {
 355                         var f *os.File
 356                         f, err = os.Create(*outputDir + "/tmp." + seqname + ".gob")
 357                         if err != nil {
 358                                 return err
 359                         }
 360                         defer os.Remove(f.Name())
 361                         bufw := bufio.NewWriterSize(f, 1<<24)
 362                         enc := gob.NewEncoder(bufw)
 363                         tmpHGVSCols[seqname] = f
 364                         todo := make(chan hgvsColSet, 128)
 365                         encodeHGVSTodo[seqname] = todo
 366                         encodeHGVS.Go(func() error {
 367                                 for colset := range todo {
 368                                         err := enc.Encode(colset)
 369                                         if err != nil {
 370                                                 encodeHGVS.Report(err)
 371                                                 for range todo {
 372                                                 }
 373                                                 return err
 374                                         }
 375                                 }
 376                                 return bufw.Flush()
 377                         })
 378                 }
 379         }
 380
 381         var toMerge [][]int16
 382         if *mergeOutput || *hgvsSingle {
 383                 toMerge = make([][]int16, len(infiles))
 384         }
 385         var onehotIndirect [][2][]uint32 // [chunkIndex][axis][index]
 386         var onehotChunkSize []uint32
 387         var onehotXrefs [][]onehotXref
 388         if *onehotSingle || *onlyPCA {
 389                 onehotIndirect = make([][2][]uint32, len(infiles))
 390                 onehotChunkSize = make([]uint32, len(infiles))
 391                 onehotXrefs = make([][]onehotXref, len(infiles))
 392         }
 393         chunkStartTag := make([]tagID, len(infiles))
 394
 395         throttleMem := throttle{Max: cmd.threads} // TODO: estimate using mem and data size
 396         throttleNumpyMem := throttle{Max: cmd.threads/2 + 1}
 397         log.Info("generating annotations and numpy matrix for each slice")
 398         var errSkip = errors.New("skip infile")
 399         var done int64
 400         for infileIdx, infile := range infiles {
 401                 infileIdx, infile := infileIdx, infile
 402                 throttleMem.Go(func() error {
 403                         seq := make(map[tagID][]TileVariant, 50000)
 404                         cgs := make(map[string]CompactGenome, len(cmd.cgnames))
 405                         f, err := open(infile)
 406                         if err != nil {
 407                                 return err
 408                         }
 409                         defer f.Close()
 410                         log.Infof("%04d: reading %s", infileIdx, infile)
 411                         err = DecodeLibrary(f, strings.HasSuffix(infile, ".gz"), func(ent *LibraryEntry) error {
 412                                 for _, tv := range ent.TileVariants {
 413                                         if tv.Ref {
 414                                                 continue
 415                                         }
 416                                         // Skip tile with no
 417                                         // corresponding ref tile, if
 418                                         // mask is in play (we can't
 419                                         // determine coordinates for
 420                                         // these)
 421                                         if mask != nil && reftile[tv.Tag] == nil {
 422                                                 continue
 423                                         }
 424                                         // Skip tile whose
 425                                         // corresponding ref tile is
 426                                         // outside target regions --
 427                                         // unless it's a potential
 428                                         // spanning tile.
 429                                         if mask != nil && reftile[tv.Tag].excluded &&
 430                                                 (int(tv.Tag+1) >= len(tagset) ||
 431                                                         (bytes.HasSuffix(tv.Sequence, tagset[tv.Tag+1]) && reftile[tv.Tag+1] != nil && !reftile[tv.Tag+1].excluded)) {
 432                                                 continue
 433                                         }
 434                                         if tv.Tag == cmd.debugTag {
 435                                                 log.Printf("infile %d %s tag %d variant %d hash %x", infileIdx, infile, tv.Tag, tv.Variant, tv.Blake2b[:3])
 436                                         }
 437                                         variants := seq[tv.Tag]
 438                                         if len(variants) == 0 {
 439                                                 variants = make([]TileVariant, 100)
 440                                         }
 441                                         for len(variants) <= int(tv.Variant) {
 442                                                 variants = append(variants, TileVariant{})
 443                                         }
 444                                         variants[int(tv.Variant)] = tv
 445                                         seq[tv.Tag] = variants
 446                                 }
 447                                 for _, cg := range ent.CompactGenomes {
 448                                         if cmd.filter.MaxTag >= 0 && cg.StartTag > tagID(cmd.filter.MaxTag) {
 449                                                 return errSkip
 450                                         }
 451                                         if !matchGenome.MatchString(cg.Name) {
 452                                                 continue
 453                                         }
 454                                         // pad to full slice size
 455                                         // to avoid out-of-bounds
 456                                         // checks later
 457                                         if sliceSize := 2 * int(cg.EndTag-cg.StartTag); len(cg.Variants) < sliceSize {
 458                                                 cg.Variants = append(cg.Variants, make([]tileVariantID, sliceSize-len(cg.Variants))...)
 459                                         }
 460                                         cgs[cg.Name] = cg
 461                                 }
 462                                 return nil
 463                         })
 464                         if err == errSkip {
 465                                 return nil
 466                         } else if err != nil {
 467                                 return fmt.Errorf("%04d: DecodeLibrary(%s): err", infileIdx, infile)
 468                         }
 469                         tagstart := cgs[cmd.cgnames[0]].StartTag
 470                         tagend := cgs[cmd.cgnames[0]].EndTag
 471                         chunkStartTag[infileIdx] = tagstart
 472
 473                         // TODO: filters
 474
 475                         log.Infof("%04d: renumber/dedup variants for tags %d-%d", infileIdx, tagstart, tagend)
 476                         variantRemap := make([][]tileVariantID, tagend-tagstart)
 477                         throttleCPU := throttle{Max: runtime.GOMAXPROCS(0)}
 478                         for tag, variants := range seq {
 479                                 tag, variants := tag, variants
 480                                 throttleCPU.Go(func() error {
 481                                         alleleCoverage := 0
 482                                         count := make(map[[blake2b.Size256]byte]int, len(variants))
 483
 484                                         rt := reftile[tag]
 485                                         if rt != nil {
 486                                                 count[blake2b.Sum256(rt.tiledata)] = 0
 487                                         }
 488
 489                                         for cgname, cg := range cgs {
 490                                                 idx := int(tag-tagstart) * 2
 491                                                 for allele := 0; allele < 2; allele++ {
 492                                                         v := cg.Variants[idx+allele]
 493                                                         if v > 0 && len(variants[v].Sequence) > 0 {
 494                                                                 count[variants[v].Blake2b]++
 495                                                                 alleleCoverage++
 496                                                         }
 497                                                         if v > 0 && tag == cmd.debugTag {
 498                                                                 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])
 499                                                         }
 500                                                 }
 501                                         }
 502                                         if alleleCoverage < cmd.minCoverage*2 {
 503                                                 idx := int(tag-tagstart) * 2
 504                                                 for _, cg := range cgs {
 505                                                         cg.Variants[idx] = 0
 506                                                         cg.Variants[idx+1] = 0
 507                                                 }
 508                                                 if tag == cmd.debugTag {
 509                                                         log.Printf("tag %d alleleCoverage %d < min %d, sample data wiped", tag, alleleCoverage, cmd.minCoverage*2)
 510                                                 }
 511                                                 return nil
 512                                         }
 513
 514                                         // hash[i] will be the hash of
 515                                         // the variant(s) that should
 516                                         // be at rank i (0-based).
 517                                         hash := make([][blake2b.Size256]byte, 0, len(count))
 518                                         for b := range count {
 519                                                 hash = append(hash, b)
 520                                         }
 521                                         sort.Slice(hash, func(i, j int) bool {
 522                                                 bi, bj := &hash[i], &hash[j]
 523                                                 if ci, cj := count[*bi], count[*bj]; ci != cj {
 524                                                         return ci > cj
 525                                                 } else {
 526                                                         return bytes.Compare((*bi)[:], (*bj)[:]) < 0
 527                                                 }
 528                                         })
 529                                         // rank[b] will be the 1-based
 530                                         // new variant number for
 531                                         // variants whose hash is b.
 532                                         rank := make(map[[blake2b.Size256]byte]tileVariantID, len(hash))
 533                                         for i, h := range hash {
 534                                                 rank[h] = tileVariantID(i + 1)
 535                                         }
 536                                         if tag == cmd.debugTag {
 537                                                 for h, r := range rank {
 538                                                         log.Printf("tag %d rank(%x) = %v", tag, h[:3], r)
 539                                                 }
 540                                         }
 541                                         // remap[v] will be the new
 542                                         // variant number for original
 543                                         // variant number v.
 544                                         remap := make([]tileVariantID, len(variants))
 545                                         for i, tv := range variants {
 546                                                 remap[i] = rank[tv.Blake2b]
 547                                         }
 548                                         if tag == cmd.debugTag {
 549                                                 for in, out := range remap {
 550                                                         if out > 0 {
 551                                                                 log.Printf("tag %d remap %d => %d", tag, in, out)
 552                                                         }
 553                                                 }
 554                                         }
 555                                         variantRemap[tag-tagstart] = remap
 556                                         if rt != nil {
 557                                                 refrank := rank[blake2b.Sum256(rt.tiledata)]
 558                                                 if tag == cmd.debugTag {
 559                                                         log.Printf("tag %d reftile variant %d => %d", tag, rt.variant, refrank)
 560                                                 }
 561                                                 rt.variant = refrank
 562                                         }
 563                                         return nil
 564                                 })
 565                         }
 566                         throttleCPU.Wait()
 567
 568                         var onehotChunk [][]int8
 569                         var onehotXref []onehotXref
 570
 571                         var annotationsFilename string
 572                         if *onlyPCA {
 573                                 annotationsFilename = "/dev/null"
 574                         } else {
 575                                 annotationsFilename = fmt.Sprintf("%s/matrix.%04d.annotations.csv", *outputDir, infileIdx)
 576                                 log.Infof("%04d: writing %s", infileIdx, annotationsFilename)
 577                         }
 578                         annof, err := os.Create(annotationsFilename)
 579                         if err != nil {
 580                                 return err
 581                         }
 582                         annow := bufio.NewWriterSize(annof, 1<<20)
 583                         outcol := 0
 584                         for tag := tagstart; tag < tagend; tag++ {
 585                                 rt := reftile[tag]
 586                                 if rt == nil && mask != nil {
 587                                         // With no ref tile, we don't
 588                                         // have coordinates to say
 589                                         // this is in the desired
 590                                         // regions -- so it's not.
 591                                         // TODO: handle ref spanning
 592                                         // tile case.
 593                                         continue
 594                                 }
 595                                 if rt != nil && rt.excluded {
 596                                         // TODO: don't skip yet --
 597                                         // first check for spanning
 598                                         // tile variants that
 599                                         // intersect non-excluded ref
 600                                         // tiles.
 601                                         continue
 602                                 }
 603                                 if cmd.filter.MaxTag >= 0 && tag > tagID(cmd.filter.MaxTag) {
 604                                         break
 605                                 }
 606                                 remap := variantRemap[tag-tagstart]
 607                                 maxv := tileVariantID(0)
 608                                 for _, v := range remap {
 609                                         if maxv < v {
 610                                                 maxv = v
 611                                         }
 612                                 }
 613                                 if *onehotChunked || *onehotSingle || *onlyPCA {
 614                                         onehot, xrefs := cmd.tv2homhet(cgs, maxv, remap, tag, tagstart, seq)
 615                                         if tag == cmd.debugTag {
 616                                                 log.WithFields(logrus.Fields{
 617                                                         "onehot": onehot,
 618                                                         "xrefs":  xrefs,
 619                                                 }).Info("tv2homhet()")
 620                                         }
 621                                         onehotChunk = append(onehotChunk, onehot...)
 622                                         onehotXref = append(onehotXref, xrefs...)
 623                                 }
 624                                 if *onlyPCA {
 625                                         outcol++
 626                                         continue
 627                                 }
 628                                 if rt == nil {
 629                                         // Reference does not use any
 630                                         // variant of this tile
 631                                         //
 632                                         // TODO: diff against the
 633                                         // relevant portion of the
 634                                         // ref's spanning tile
 635                                         outcol++
 636                                         continue
 637                                 }
 638                                 fmt.Fprintf(annow, "%d,%d,%d,=,%s,%d,,,\n", tag, outcol, rt.variant, rt.seqname, rt.pos)
 639                                 variants := seq[tag]
 640                                 reftilestr := strings.ToUpper(string(rt.tiledata))
 641
 642                                 done := make([]bool, maxv+1)
 643                                 variantDiffs := make([][]hgvs.Variant, maxv+1)
 644                                 for v, tv := range variants {
 645                                         v := remap[v]
 646                                         if v == 0 || v == rt.variant || done[v] {
 647                                                 continue
 648                                         } else {
 649                                                 done[v] = true
 650                                         }
 651                                         if len(tv.Sequence) < taglen {
 652                                                 continue
 653                                         }
 654                                         // if reftilestr doesn't end
 655                                         // in the same tag as tv,
 656                                         // extend reftilestr with
 657                                         // following ref tiles until
 658                                         // it does (up to an arbitrary
 659                                         // sanity-check limit)
 660                                         reftilestr := reftilestr
 661                                         endtagstr := strings.ToUpper(string(tv.Sequence[len(tv.Sequence)-taglen:]))
 662                                         for i, rt := 0, rt; i < annotationMaxTileSpan && !strings.HasSuffix(reftilestr, endtagstr) && rt.nexttag >= 0; i++ {
 663                                                 rt = reftile[rt.nexttag]
 664                                                 if rt == nil {
 665                                                         break
 666                                                 }
 667                                                 reftilestr += strings.ToUpper(string(rt.tiledata[taglen:]))
 668                                         }
 669                                         if mask != nil && !mask.Check(strings.TrimPrefix(rt.seqname, "chr"), rt.pos, rt.pos+len(reftilestr)) {
 670                                                 continue
 671                                         }
 672                                         if !strings.HasSuffix(reftilestr, endtagstr) {
 673                                                 fmt.Fprintf(annow, "%d,%d,%d,,%s,%d,,,\n", tag, outcol, v, rt.seqname, rt.pos)
 674                                                 continue
 675                                         }
 676                                         if lendiff := len(reftilestr) - len(tv.Sequence); lendiff < -1000 || lendiff > 1000 {
 677                                                 fmt.Fprintf(annow, "%d,%d,%d,,%s,%d,,,\n", tag, outcol, v, rt.seqname, rt.pos)
 678                                                 continue
 679                                         }
 680                                         diffs, _ := hgvs.Diff(reftilestr, strings.ToUpper(string(tv.Sequence)), 0)
 681                                         for i := range diffs {
 682                                                 diffs[i].Position += rt.pos
 683                                         }
 684                                         for _, diff := range diffs {
 685                                                 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)
 686                                         }
 687                                         if *hgvsChunked {
 688                                                 variantDiffs[v] = diffs
 689                                         }
 690                                 }
 691                                 if *hgvsChunked {
 692                                         // We can now determine, for each HGVS
 693                                         // variant (diff) in this reftile
 694                                         // region, whether a given genome
 695                                         // phase/allele (1) has the variant, (0) has
 696                                         // =ref or a different variant in that
 697                                         // position, or (-1) is lacking
 698                                         // coverage / couldn't be diffed.
 699                                         hgvsCol := hgvsColSet{}
 700                                         for _, diffs := range variantDiffs {
 701                                                 for _, diff := range diffs {
 702                                                         if _, ok := hgvsCol[diff]; ok {
 703                                                                 continue
 704                                                         }
 705                                                         hgvsCol[diff] = [2][]int8{
 706                                                                 make([]int8, len(cmd.cgnames)),
 707                                                                 make([]int8, len(cmd.cgnames)),
 708                                                         }
 709                                                 }
 710                                         }
 711                                         for row, name := range cmd.cgnames {
 712                                                 variants := cgs[name].Variants[(tag-tagstart)*2:]
 713                                                 for ph := 0; ph < 2; ph++ {
 714                                                         v := variants[ph]
 715                                                         if int(v) >= len(remap) {
 716                                                                 v = 0
 717                                                         } else {
 718                                                                 v = remap[v]
 719                                                         }
 720                                                         if v == rt.variant {
 721                                                                 // hgvsCol[*][ph][row] is already 0
 722                                                         } else if len(variantDiffs[v]) == 0 {
 723                                                                 // lacking coverage / couldn't be diffed
 724                                                                 for _, col := range hgvsCol {
 725                                                                         col[ph][row] = -1
 726                                                                 }
 727                                                         } else {
 728                                                                 for _, diff := range variantDiffs[v] {
 729                                                                         hgvsCol[diff][ph][row] = 1
 730                                                                 }
 731                                                         }
 732                                                 }
 733                                         }
 734                                         for diff, colpair := range hgvsCol {
 735                                                 allele2homhet(colpair)
 736                                                 if !cmd.filterHGVScolpair(colpair) {
 737                                                         delete(hgvsCol, diff)
 738                                                 }
 739                                         }
 740                                         if len(hgvsCol) > 0 {
 741                                                 encodeHGVSTodo[rt.seqname] <- hgvsCol
 742                                         }
 743                                 }
 744                                 outcol++
 745                         }
 746                         err = annow.Flush()
 747                         if err != nil {
 748                                 return err
 749                         }
 750                         err = annof.Close()
 751                         if err != nil {
 752                                 return err
 753                         }
 754
 755                         if *onehotChunked {
 756                                 // transpose onehotChunk[col][row] to numpy[row*ncols+col]
 757                                 rows := len(cmd.cgnames)
 758                                 cols := len(onehotChunk)
 759                                 log.Infof("%04d: preparing onehot numpy (rows=%d, cols=%d, mem=%d)", infileIdx, rows, cols, rows*cols)
 760                                 throttleNumpyMem.Acquire()
 761                                 out := onehotcols2int8(onehotChunk)
 762                                 fnm := fmt.Sprintf("%s/onehot.%04d.npy", *outputDir, infileIdx)
 763                                 err = writeNumpyInt8(fnm, out, rows, cols)
 764                                 if err != nil {
 765                                         return err
 766                                 }
 767                                 fnm = fmt.Sprintf("%s/onehot-columns.%04d.npy", *outputDir, infileIdx)
 768                                 err = writeNumpyInt32(fnm, onehotXref2int32(onehotXref), 4, len(onehotXref))
 769                                 if err != nil {
 770                                         return err
 771                                 }
 772                                 debug.FreeOSMemory()
 773                                 throttleNumpyMem.Release()
 774                         }
 775                         if *onehotSingle || *onlyPCA {
 776                                 onehotIndirect[infileIdx] = onehotChunk2Indirect(onehotChunk)
 777                                 onehotChunkSize[infileIdx] = uint32(len(onehotChunk))
 778                                 onehotXrefs[infileIdx] = onehotXref
 779                                 n := len(onehotIndirect[infileIdx][0])
 780                                 log.Infof("%04d: keeping onehot coordinates in memory (n=%d, mem=%d)", infileIdx, n, n*8*2)
 781                         }
 782                         if !(*onehotSingle || *onehotChunked || *onlyPCA) || *mergeOutput || *hgvsSingle {
 783                                 log.Infof("%04d: preparing numpy (rows=%d, cols=%d)", infileIdx, len(cmd.cgnames), 2*outcol)
 784                                 throttleNumpyMem.Acquire()
 785                                 rows := len(cmd.cgnames)
 786                                 cols := 2 * outcol
 787                                 out := make([]int16, rows*cols)
 788                                 for row, name := range cmd.cgnames {
 789                                         outidx := row * cols
 790                                         for col, v := range cgs[name].Variants {
 791                                                 tag := tagstart + tagID(col/2)
 792                                                 if cmd.filter.MaxTag >= 0 && tag > tagID(cmd.filter.MaxTag) {
 793                                                         break
 794                                                 }
 795                                                 if rt := reftile[tag]; rt == nil || rt.excluded {
 796                                                         continue
 797                                                 }
 798                                                 if v == 0 {
 799                                                         out[outidx] = 0 // tag not found / spanning tile
 800                                                 } else if variants, ok := seq[tag]; ok && int(v) < len(variants) && len(variants[v].Sequence) > 0 {
 801                                                         out[outidx] = int16(variantRemap[tag-tagstart][v])
 802                                                 } else {
 803                                                         out[outidx] = -1 // low quality tile variant
 804                                                 }
 805                                                 if tag == cmd.debugTag {
 806                                                         log.Printf("tag %d row %d col %d outidx %d v %d out %d", tag, row, col, outidx, v, out[outidx])
 807                                                 }
 808                                                 outidx++
 809                                         }
 810                                 }
 811                                 seq = nil
 812                                 cgs = nil
 813                                 debug.FreeOSMemory()
 814                                 throttleNumpyMem.Release()
 815                                 if *mergeOutput || *hgvsSingle {
 816                                         log.Infof("%04d: matrix fragment %d rows x %d cols", infileIdx, rows, cols)
 817                                         toMerge[infileIdx] = out
 818                                 }
 819                                 if !*mergeOutput && !*onehotChunked && !*onehotSingle {
 820                                         fnm := fmt.Sprintf("%s/matrix.%04d.npy", *outputDir, infileIdx)
 821                                         err = writeNumpyInt16(fnm, out, rows, cols)
 822                                         if err != nil {
 823                                                 return err
 824                                         }
 825                                 }
 826                         }
 827                         debug.FreeOSMemory()
 828                         log.Infof("%s: done (%d/%d)", infile, int(atomic.AddInt64(&done, 1)), len(infiles))
 829                         return nil
 830                 })
 831         }
 832         if err = throttleMem.Wait(); err != nil {
 833                 return err
 834         }
 835
 836         if *hgvsChunked {
 837                 log.Info("flushing hgvsCols temp files")
 838                 for seqname := range refseq {
 839                         close(encodeHGVSTodo[seqname])
 840                 }
 841                 err = encodeHGVS.Wait()
 842                 if err != nil {
 843                         return err
 844                 }
 845                 for seqname := range refseq {
 846                         log.Infof("%s: reading hgvsCols from temp file", seqname)
 847                         f := tmpHGVSCols[seqname]
 848                         _, err = f.Seek(0, io.SeekStart)
 849                         if err != nil {
 850                                 return err
 851                         }
 852                         var hgvsCols hgvsColSet
 853                         dec := gob.NewDecoder(bufio.NewReaderSize(f, 1<<24))
 854                         for err == nil {
 855                                 err = dec.Decode(&hgvsCols)
 856                         }
 857                         if err != io.EOF {
 858                                 return err
 859                         }
 860                         log.Infof("%s: sorting %d hgvs variants", seqname, len(hgvsCols))
 861                         variants := make([]hgvs.Variant, 0, len(hgvsCols))
 862                         for v := range hgvsCols {
 863                                 variants = append(variants, v)
 864                         }
 865                         sort.Slice(variants, func(i, j int) bool {
 866                                 vi, vj := &variants[i], &variants[j]
 867                                 if vi.Position != vj.Position {
 868                                         return vi.Position < vj.Position
 869                                 } else if vi.Ref != vj.Ref {
 870                                         return vi.Ref < vj.Ref
 871                                 } else {
 872                                         return vi.New < vj.New
 873                                 }
 874                         })
 875                         rows := len(cmd.cgnames)
 876                         cols := len(variants) * 2
 877                         log.Infof("%s: building hgvs matrix (rows=%d, cols=%d, mem=%d)", seqname, rows, cols, rows*cols)
 878                         out := make([]int8, rows*cols)
 879                         for varIdx, variant := range variants {
 880                                 hgvsCols := hgvsCols[variant]
 881                                 for row := range cmd.cgnames {
 882                                         for ph := 0; ph < 2; ph++ {
 883                                                 out[row*cols+varIdx+ph] = hgvsCols[ph][row]
 884                                         }
 885                                 }
 886                         }
 887                         err = writeNumpyInt8(fmt.Sprintf("%s/hgvs.%s.npy", *outputDir, seqname), out, rows, cols)
 888                         if err != nil {
 889                                 return err
 890                         }
 891                         out = nil
 892
 893                         fnm := fmt.Sprintf("%s/hgvs.%s.annotations.csv", *outputDir, seqname)
 894                         log.Infof("%s: writing hgvs column labels to %s", seqname, fnm)
 895                         var hgvsLabels bytes.Buffer
 896                         for varIdx, variant := range variants {
 897                                 fmt.Fprintf(&hgvsLabels, "%d,%s:g.%s\n", varIdx, seqname, variant.String())
 898                         }
 899                         err = ioutil.WriteFile(fnm, hgvsLabels.Bytes(), 0666)
 900                         if err != nil {
 901                                 return err
 902                         }
 903                 }
 904         }
 905
 906         if *mergeOutput || *hgvsSingle {
 907                 var annow *bufio.Writer
 908                 var annof *os.File
 909                 if *mergeOutput {
 910                         annoFilename := fmt.Sprintf("%s/matrix.annotations.csv", *outputDir)
 911                         annof, err = os.Create(annoFilename)
 912                         if err != nil {
 913                                 return err
 914                         }
 915                         annow = bufio.NewWriterSize(annof, 1<<20)
 916                 }
 917
 918                 rows := len(cmd.cgnames)
 919                 cols := 0
 920                 for _, chunk := range toMerge {
 921                         cols += len(chunk) / rows
 922                 }
 923                 log.Infof("merging output matrix (rows=%d, cols=%d, mem=%d) and annotations", rows, cols, rows*cols*2)
 924                 var out []int16
 925                 if *mergeOutput {
 926                         out = make([]int16, rows*cols)
 927                 }
 928                 hgvsCols := map[string][2][]int16{} // hgvs -> [[g0,g1,g2,...], [g0,g1,g2,...]] (slice of genomes for each phase)
 929                 startcol := 0
 930                 for outIdx, chunk := range toMerge {
 931                         chunkcols := len(chunk) / rows
 932                         if *mergeOutput {
 933                                 for row := 0; row < rows; row++ {
 934                                         copy(out[row*cols+startcol:], chunk[row*chunkcols:(row+1)*chunkcols])
 935                                 }
 936                         }
 937                         toMerge[outIdx] = nil
 938
 939                         annotationsFilename := fmt.Sprintf("%s/matrix.%04d.annotations.csv", *outputDir, outIdx)
 940                         log.Infof("reading %s", annotationsFilename)
 941                         buf, err := os.ReadFile(annotationsFilename)
 942                         if err != nil {
 943                                 return err
 944                         }
 945                         if *mergeOutput {
 946                                 err = os.Remove(annotationsFilename)
 947                                 if err != nil {
 948                                         return err
 949                                 }
 950                         }
 951                         for _, line := range bytes.Split(buf, []byte{'\n'}) {
 952                                 if len(line) == 0 {
 953                                         continue
 954                                 }
 955                                 fields := bytes.SplitN(line, []byte{','}, 9)
 956                                 tag, _ := strconv.Atoi(string(fields[0]))
 957                                 incol, _ := strconv.Atoi(string(fields[1]))
 958                                 tileVariant, _ := strconv.Atoi(string(fields[2]))
 959                                 hgvsID := string(fields[3])
 960                                 seqname := string(fields[4])
 961                                 pos, _ := strconv.Atoi(string(fields[5]))
 962                                 refseq := fields[6]
 963                                 if hgvsID == "" {
 964                                         // Null entry for un-diffable
 965                                         // tile variant
 966                                         continue
 967                                 }
 968                                 if hgvsID == "=" {
 969                                         // Null entry for ref tile
 970                                         continue
 971                                 }
 972                                 if mask != nil && !mask.Check(strings.TrimPrefix(seqname, "chr"), pos, pos+len(refseq)) {
 973                                         // The tile intersects one of
 974                                         // the selected regions, but
 975                                         // this particular HGVS
 976                                         // variant does not.
 977                                         continue
 978                                 }
 979                                 hgvsColPair := hgvsCols[hgvsID]
 980                                 if hgvsColPair[0] == nil {
 981                                         // values in new columns start
 982                                         // out as -1 ("no data yet")
 983                                         // or 0 ("=ref") here, may
 984                                         // change to 1 ("hgvs variant
 985                                         // present") below, either on
 986                                         // this line or a future line.
 987                                         hgvsColPair = [2][]int16{make([]int16, len(cmd.cgnames)), make([]int16, len(cmd.cgnames))}
 988                                         rt, ok := reftile[tagID(tag)]
 989                                         if !ok {
 990                                                 err = fmt.Errorf("bug: seeing annotations for tag %d, but it has no reftile entry", tag)
 991                                                 return err
 992                                         }
 993                                         for ph := 0; ph < 2; ph++ {
 994                                                 for row := 0; row < rows; row++ {
 995                                                         v := chunk[row*chunkcols+incol*2+ph]
 996                                                         if tileVariantID(v) == rt.variant {
 997                                                                 hgvsColPair[ph][row] = 0
 998                                                         } else {
 999                                                                 hgvsColPair[ph][row] = -1
1000                                                         }
1001                                                 }
1002                                         }
1003                                         hgvsCols[hgvsID] = hgvsColPair
1004                                         if annow != nil {
1005                                                 hgvsref := hgvs.Variant{
1006                                                         Position: pos,
1007                                                         Ref:      string(refseq),
1008                                                         New:      string(refseq),
1009                                                 }
1010                                                 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])
1011                                         }
1012                                 }
1013                                 if annow != nil {
1014                                         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])
1015                                 }
1016                                 for ph := 0; ph < 2; ph++ {
1017                                         for row := 0; row < rows; row++ {
1018                                                 v := chunk[row*chunkcols+incol*2+ph]
1019                                                 if int(v) == tileVariant {
1020                                                         hgvsColPair[ph][row] = 1
1021                                                 }
1022                                         }
1023                                 }
1024                         }
1025
1026                         startcol += chunkcols
1027                 }
1028                 if *mergeOutput {
1029                         err = annow.Flush()
1030                         if err != nil {
1031                                 return err
1032                         }
1033                         err = annof.Close()
1034                         if err != nil {
1035                                 return err
1036                         }
1037                         err = writeNumpyInt16(fmt.Sprintf("%s/matrix.npy", *outputDir), out, rows, cols)
1038                         if err != nil {
1039                                 return err
1040                         }
1041                 }
1042                 out = nil
1043
1044                 if *hgvsSingle {
1045                         cols = len(hgvsCols) * 2
1046                         log.Printf("building hgvs-based matrix: %d rows x %d cols", rows, cols)
1047                         out = make([]int16, rows*cols)
1048                         hgvsIDs := make([]string, 0, cols/2)
1049                         for hgvsID := range hgvsCols {
1050                                 hgvsIDs = append(hgvsIDs, hgvsID)
1051                         }
1052                         sort.Strings(hgvsIDs)
1053                         var hgvsLabels bytes.Buffer
1054                         for idx, hgvsID := range hgvsIDs {
1055                                 fmt.Fprintf(&hgvsLabels, "%d,%s\n", idx, hgvsID)
1056                                 for ph := 0; ph < 2; ph++ {
1057                                         hgvscol := hgvsCols[hgvsID][ph]
1058                                         for row, val := range hgvscol {
1059                                                 out[row*cols+idx*2+ph] = val
1060                                         }
1061                                 }
1062                         }
1063                         err = writeNumpyInt16(fmt.Sprintf("%s/hgvs.npy", *outputDir), out, rows, cols)
1064                         if err != nil {
1065                                 return err
1066                         }
1067
1068                         fnm := fmt.Sprintf("%s/hgvs.annotations.csv", *outputDir)
1069                         log.Printf("writing hgvs labels: %s", fnm)
1070                         err = ioutil.WriteFile(fnm, hgvsLabels.Bytes(), 0777)
1071                         if err != nil {
1072                                 return err
1073                         }
1074                 }
1075         }
1076         if *onehotSingle || *onlyPCA {
1077                 nzCount := 0
1078                 for _, part := range onehotIndirect {
1079                         nzCount += len(part[0])
1080                 }
1081                 onehot := make([]uint32, nzCount*2) // [r,r,r,...,c,c,c,...]
1082                 var xrefs []onehotXref
1083                 chunkOffset := uint32(0)
1084                 outcol := 0
1085                 for i, part := range onehotIndirect {
1086                         for i := range part[1] {
1087                                 part[1][i] += chunkOffset
1088                         }
1089                         copy(onehot[outcol:], part[0])
1090                         copy(onehot[outcol+nzCount:], part[1])
1091                         xrefs = append(xrefs, onehotXrefs[i]...)
1092
1093                         outcol += len(part[0])
1094                         chunkOffset += onehotChunkSize[i]
1095
1096                         part[0] = nil
1097                         part[1] = nil
1098                         onehotXrefs[i] = nil
1099                         debug.FreeOSMemory()
1100                 }
1101                 if *onehotSingle {
1102                         fnm := fmt.Sprintf("%s/onehot.npy", *outputDir)
1103                         err = writeNumpyUint32(fnm, onehot, 2, nzCount)
1104                         if err != nil {
1105                                 return err
1106                         }
1107                         fnm = fmt.Sprintf("%s/onehot-columns.npy", *outputDir)
1108                         err = writeNumpyInt32(fnm, onehotXref2int32(xrefs), 5, len(xrefs))
1109                         if err != nil {
1110                                 return err
1111                         }
1112                 }
1113                 if *onlyPCA {
1114                         cols := 0
1115                         for _, c := range onehot[nzCount:] {
1116                                 if int(c) >= cols {
1117                                         cols = int(c) + 1
1118                                 }
1119                         }
1120                         if cols == 0 {
1121                                 return fmt.Errorf("cannot do PCA: one-hot matrix is empty")
1122                         }
1123                         log.Printf("have %d one-hot cols", cols)
1124                         stride := 1
1125                         for *maxPCATiles > 0 && cols > *maxPCATiles*2 {
1126                                 cols = (cols + 1) / 2
1127                                 stride = stride * 2
1128                         }
1129                         log.Printf("creating full matrix (%d rows) and training matrix (%d rows) with %d cols, stride %d", len(cmd.cgnames), cmd.trainingSetSize, cols, stride)
1130                         mtxFull := mat.NewDense(len(cmd.cgnames), cols, nil)
1131                         mtxTrain := mat.NewDense(cmd.trainingSetSize, cols, nil)
1132                         for i, c := range onehot[nzCount:] {
1133                                 if int(c/2)%stride == 0 {
1134                                         outcol := int(c/2)/stride*2 + int(c)%2
1135                                         mtxFull.Set(int(onehot[i]), outcol, 1)
1136                                         if trainRow := cmd.trainingSet[int(onehot[i])]; trainRow >= 0 {
1137                                                 mtxTrain.Set(trainRow, outcol, 1)
1138                                         }
1139                                 }
1140                         }
1141                         log.Print("fitting")
1142                         transformer := nlp.NewPCA(*pcaComponents)
1143                         transformer.Fit(mtxTrain.T())
1144                         log.Printf("transforming")
1145                         pca, err := transformer.Transform(mtxFull.T())
1146                         if err != nil {
1147                                 return err
1148                         }
1149                         pca = pca.T()
1150                         outrows, outcols := pca.Dims()
1151                         log.Printf("copying result to numpy output array: %d rows, %d cols", outrows, outcols)
1152                         out := make([]float64, outrows*outcols)
1153                         for i := 0; i < outrows; i++ {
1154                                 for j := 0; j < outcols; j++ {
1155                                         out[i*outcols+j] = pca.At(i, j)
1156                                 }
1157                         }
1158                         fnm := fmt.Sprintf("%s/pca.npy", *outputDir)
1159                         log.Printf("writing numpy: %s", fnm)
1160                         output, err := os.OpenFile(fnm, os.O_CREATE|os.O_TRUNC|os.O_WRONLY, 0777)
1161                         if err != nil {
1162                                 return err
1163                         }
1164                         npw, err := gonpy.NewWriter(nopCloser{output})
1165                         if err != nil {
1166                                 return fmt.Errorf("gonpy.NewWriter: %w", err)
1167                         }
1168                         npw.Shape = []int{outrows, outcols}
1169                         err = npw.WriteFloat64(out)
1170                         if err != nil {
1171                                 return fmt.Errorf("WriteFloat64: %w", err)
1172                         }
1173                         err = output.Close()
1174                         if err != nil {
1175                                 return err
1176                         }
1177                         log.Print("done")
1178
1179                         samplesOutFilename := *outputDir + "/samples.csv"
1180                         log.Infof("writing sample metadata to %s", samplesOutFilename)
1181                         var f *os.File
1182                         f, err = os.Create(samplesOutFilename)
1183                         if err != nil {
1184                                 return err
1185                         }
1186                         defer f.Close()
1187                         for i, si := range cmd.samples {
1188                                 var cc, tv string
1189                                 if si.isCase {
1190                                         cc = "1"
1191                                 } else if si.isControl {
1192                                         cc = "0"
1193                                 }
1194                                 if si.isTraining {
1195                                         tv = "1"
1196                                 } else {
1197                                         tv = "0"
1198                                 }
1199                                 var pcavals string
1200                                 for c := 0; c < outcols; c++ {
1201                                         pcavals += fmt.Sprintf(",%f", pca.At(i, c))
1202                                 }
1203                                 _, err = fmt.Fprintf(f, "%d,%s,%s,%s%s\n", i, si.id, cc, tv, pcavals)
1204                                 if err != nil {
1205                                         err = fmt.Errorf("write %s: %w", samplesOutFilename, err)
1206                                         return err
1207                                 }
1208                         }
1209                         err = f.Close()
1210                         if err != nil {
1211                                 err = fmt.Errorf("close %s: %w", samplesOutFilename, err)
1212                                 return err
1213                         }
1214                         log.Print("done")
1215                 }
1216         }
1217         if !*mergeOutput && !*onehotChunked && !*onehotSingle && !*onlyPCA {
1218                 tagoffsetFilename := *outputDir + "/chunk-tag-offset.csv"
1219                 log.Infof("writing tag offsets to %s", tagoffsetFilename)
1220                 var f *os.File
1221                 f, err = os.Create(tagoffsetFilename)
1222                 if err != nil {
1223                         return err
1224                 }
1225                 defer f.Close()
1226                 for idx, offset := range chunkStartTag {
1227                         _, err = fmt.Fprintf(f, "%q,%d\n", fmt.Sprintf("matrix.%04d.npy", idx), offset)
1228                         if err != nil {
1229                                 err = fmt.Errorf("write %s: %w", tagoffsetFilename, err)
1230                                 return err
1231                         }
1232                 }
1233                 err = f.Close()
1234                 if err != nil {
1235                         err = fmt.Errorf("close %s: %w", tagoffsetFilename, err)
1236                         return err
1237                 }
1238         }
1239
1240         return nil
1241 }
1242
1243 type sampleInfo struct {
1244         id            string
1245         isCase        bool
1246         isControl     bool
1247         isTraining    bool
1248         isValidation  bool
1249         pcaComponents []float64
1250 }
1251
1252 // Read samples.csv file with case/control and training/validation
1253 // flags.
1254 func (cmd *sliceNumpy) loadSampleInfo(samplesFilename string) ([]sampleInfo, error) {
1255         var si []sampleInfo
1256         f, err := open(samplesFilename)
1257         if err != nil {
1258                 return nil, err
1259         }
1260         buf, err := io.ReadAll(f)
1261         f.Close()
1262         if err != nil {
1263                 return nil, err
1264         }
1265         lineNum := 0
1266         for _, csv := range bytes.Split(buf, []byte{'\n'}) {
1267                 lineNum++
1268                 split := strings.Split(string(csv), ",")
1269                 if len(split) != 4 {
1270                         return nil, fmt.Errorf("%d fields != 4 in %s line %d: %q", len(split), samplesFilename, lineNum, csv)
1271                 }
1272                 if split[0] == "Index" && split[1] == "SampleID" && split[2] == "CaseControl" && split[3] == "TrainingValidation" {
1273                         continue
1274                 }
1275                 idx, err := strconv.Atoi(split[0])
1276                 if err != nil {
1277                         if lineNum == 1 {
1278                                 return nil, fmt.Errorf("header does not look right: %q", csv)
1279                         }
1280                         return nil, fmt.Errorf("%s line %d: index: %s", samplesFilename, lineNum, err)
1281                 }
1282                 if idx != len(si) {
1283                         return nil, fmt.Errorf("%s line %d: index %d out of order", samplesFilename, lineNum, idx)
1284                 }
1285                 si = append(si, sampleInfo{
1286                         id:           split[1],
1287                         isCase:       split[2] == "1",
1288                         isControl:    split[2] == "0",
1289                         isTraining:   split[3] == "1",
1290                         isValidation: split[3] == "0",
1291                 })
1292         }
1293         return si, nil
1294 }
1295
1296 func (cmd *sliceNumpy) filterHGVScolpair(colpair [2][]int8) bool {
1297         if cmd.chi2PValue >= 1 {
1298                 return true
1299         }
1300         col0 := make([]bool, 0, len(cmd.chi2Cases))
1301         col1 := make([]bool, 0, len(cmd.chi2Cases))
1302         cases := make([]bool, 0, len(cmd.chi2Cases))
1303         for i, c := range cmd.chi2Cases {
1304                 if colpair[0][i] < 0 {
1305                         continue
1306                 }
1307                 col0 = append(col0, colpair[0][i] != 0)
1308                 col1 = append(col1, colpair[1][i] != 0)
1309                 cases = append(cases, c)
1310         }
1311         return len(cases) >= cmd.minCoverage &&
1312                 (pvalue(col0, cases) <= cmd.chi2PValue || pvalue(col1, cases) <= cmd.chi2PValue)
1313 }
1314
1315 func writeNumpyUint32(fnm string, out []uint32, rows, cols int) error {
1316         output, err := os.Create(fnm)
1317         if err != nil {
1318                 return err
1319         }
1320         defer output.Close()
1321         bufw := bufio.NewWriterSize(output, 1<<26)
1322         npw, err := gonpy.NewWriter(nopCloser{bufw})
1323         if err != nil {
1324                 return err
1325         }
1326         log.WithFields(log.Fields{
1327                 "filename": fnm,
1328                 "rows":     rows,
1329                 "cols":     cols,
1330                 "bytes":    rows * cols * 4,
1331         }).Infof("writing numpy: %s", fnm)
1332         npw.Shape = []int{rows, cols}
1333         npw.WriteUint32(out)
1334         err = bufw.Flush()
1335         if err != nil {
1336                 return err
1337         }
1338         return output.Close()
1339 }
1340
1341 func writeNumpyInt32(fnm string, out []int32, rows, cols int) error {
1342         output, err := os.Create(fnm)
1343         if err != nil {
1344                 return err
1345         }
1346         defer output.Close()
1347         bufw := bufio.NewWriterSize(output, 1<<26)
1348         npw, err := gonpy.NewWriter(nopCloser{bufw})
1349         if err != nil {
1350                 return err
1351         }
1352         log.WithFields(log.Fields{
1353                 "filename": fnm,
1354                 "rows":     rows,
1355                 "cols":     cols,
1356                 "bytes":    rows * cols * 4,
1357         }).Infof("writing numpy: %s", fnm)
1358         npw.Shape = []int{rows, cols}
1359         npw.WriteInt32(out)
1360         err = bufw.Flush()
1361         if err != nil {
1362                 return err
1363         }
1364         return output.Close()
1365 }
1366
1367 func writeNumpyInt16(fnm string, out []int16, rows, cols int) error {
1368         output, err := os.Create(fnm)
1369         if err != nil {
1370                 return err
1371         }
1372         defer output.Close()
1373         bufw := bufio.NewWriterSize(output, 1<<26)
1374         npw, err := gonpy.NewWriter(nopCloser{bufw})
1375         if err != nil {
1376                 return err
1377         }
1378         log.WithFields(log.Fields{
1379                 "filename": fnm,
1380                 "rows":     rows,
1381                 "cols":     cols,
1382                 "bytes":    rows * cols * 2,
1383         }).Infof("writing numpy: %s", fnm)
1384         npw.Shape = []int{rows, cols}
1385         npw.WriteInt16(out)
1386         err = bufw.Flush()
1387         if err != nil {
1388                 return err
1389         }
1390         return output.Close()
1391 }
1392
1393 func writeNumpyInt8(fnm string, out []int8, rows, cols int) error {
1394         output, err := os.Create(fnm)
1395         if err != nil {
1396                 return err
1397         }
1398         defer output.Close()
1399         bufw := bufio.NewWriterSize(output, 1<<26)
1400         npw, err := gonpy.NewWriter(nopCloser{bufw})
1401         if err != nil {
1402                 return err
1403         }
1404         log.WithFields(log.Fields{
1405                 "filename": fnm,
1406                 "rows":     rows,
1407                 "cols":     cols,
1408                 "bytes":    rows * cols,
1409         }).Infof("writing numpy: %s", fnm)
1410         npw.Shape = []int{rows, cols}
1411         npw.WriteInt8(out)
1412         err = bufw.Flush()
1413         if err != nil {
1414                 return err
1415         }
1416         return output.Close()
1417 }
1418
1419 func allele2homhet(colpair [2][]int8) {
1420         a, b := colpair[0], colpair[1]
1421         for i, av := range a {
1422                 bv := b[i]
1423                 if av < 0 || bv < 0 {
1424                         // no-call
1425                         a[i], b[i] = -1, -1
1426                 } else if av > 0 && bv > 0 {
1427                         // hom
1428                         a[i], b[i] = 1, 0
1429                 } else if av > 0 || bv > 0 {
1430                         // het
1431                         a[i], b[i] = 0, 1
1432                 } else {
1433                         // ref (or a different variant in same position)
1434                         // (this is a no-op) a[i], b[i] = 0, 0
1435                 }
1436         }
1437 }
1438
1439 type onehotXref struct {
1440         tag     tagID
1441         variant tileVariantID
1442         hom     bool
1443         pvalue  float64
1444 }
1445
1446 const onehotXrefSize = unsafe.Sizeof(onehotXref{})
1447
1448 // Build onehot matrix (m[tileVariantIndex][genome] == 0 or 1) for all
1449 // variants of a single tile/tag#.
1450 //
1451 // Return nil if no tile variant passes Χ² filter.
1452 func (cmd *sliceNumpy) tv2homhet(cgs map[string]CompactGenome, maxv tileVariantID, remap []tileVariantID, tag, chunkstarttag tagID, seq map[tagID][]TileVariant) ([][]int8, []onehotXref) {
1453         if tag == cmd.debugTag {
1454                 tv := make([]tileVariantID, len(cmd.cgnames)*2)
1455                 for i, name := range cmd.cgnames {
1456                         copy(tv[i*2:(i+1)*2], cgs[name].Variants[(tag-chunkstarttag)*2:])
1457                 }
1458                 log.WithFields(logrus.Fields{
1459                         "cgs[i].Variants[tag*2+j]": tv,
1460                         "maxv":                     maxv,
1461                         "remap":                    remap,
1462                         "tag":                      tag,
1463                         "chunkstarttag":            chunkstarttag,
1464                 }).Info("tv2homhet()")
1465         }
1466         if maxv < 1 || (maxv < 2 && !cmd.includeVariant1) {
1467                 // everyone has the most common variant (of the variants we don't drop)
1468                 return nil, nil
1469         }
1470         tagoffset := tag - chunkstarttag
1471         coverage := 0
1472         for _, cg := range cgs {
1473                 alleles := 0
1474                 for _, v := range cg.Variants[tagoffset*2 : tagoffset*2+2] {
1475                         if v > 0 && int(v) < len(seq[tag]) && len(seq[tag][v].Sequence) > 0 {
1476                                 alleles++
1477                         }
1478                 }
1479                 if alleles == 2 {
1480                         coverage++
1481                 }
1482         }
1483         if coverage < cmd.minCoverage {
1484                 return nil, nil
1485         }
1486         obs := make([][]bool, (maxv+1)*2) // 2 slices (hom + het) for each variant#
1487         for i := range obs {
1488                 obs[i] = make([]bool, len(cmd.cgnames))
1489         }
1490         for cgid, name := range cmd.cgnames {
1491                 cgvars := cgs[name].Variants[tagoffset*2:]
1492                 tv0, tv1 := remap[cgvars[0]], remap[cgvars[1]]
1493                 for v := tileVariantID(1); v <= maxv; v++ {
1494                         if tv0 == v && tv1 == v {
1495                                 obs[v*2][cgid] = true
1496                         } else if tv0 == v || tv1 == v {
1497                                 obs[v*2+1][cgid] = true
1498                         }
1499                 }
1500         }
1501         var onehot [][]int8
1502         var xref []onehotXref
1503         for col := 2; col < len(obs); col++ {
1504                 // col 0,1 correspond to tile variant 0, i.e.,
1505                 // no-call; col 2,3 correspond to the most common
1506                 // variant; so we (normally) start at col 4.
1507                 if col < 4 && !cmd.includeVariant1 {
1508                         continue
1509                 }
1510                 p := pvalue(obs[col], cmd.chi2Cases)
1511                 if cmd.chi2PValue < 1 && !(p < cmd.chi2PValue) {
1512                         continue
1513                 }
1514                 onehot = append(onehot, bool2int8(obs[col]))
1515                 xref = append(xref, onehotXref{
1516                         tag:     tag,
1517                         variant: tileVariantID(col >> 1),
1518                         hom:     col&1 == 0,
1519                         pvalue:  p,
1520                 })
1521         }
1522         return onehot, xref
1523 }
1524
1525 func bool2int8(in []bool) []int8 {
1526         out := make([]int8, len(in))
1527         for i, v := range in {
1528                 if v {
1529                         out[i] = 1
1530                 }
1531         }
1532         return out
1533 }
1534
1535 // convert a []onehotXref with length N to a numpy-style []int32
1536 // matrix with N columns, one row per field of onehotXref struct.
1537 //
1538 // Hom/het row contains hom=0, het=1.
1539 //
1540 // P-value row contains 1000000x actual p-value.
1541 func onehotXref2int32(xrefs []onehotXref) []int32 {
1542         xcols := len(xrefs)
1543         xdata := make([]int32, 5*xcols)
1544         for i, xref := range xrefs {
1545                 xdata[i] = int32(xref.tag)
1546                 xdata[xcols+i] = int32(xref.variant)
1547                 if xref.hom {
1548                         xdata[xcols*2+i] = 1
1549                 }
1550                 xdata[xcols*3+i] = int32(xref.pvalue * 1000000)
1551                 xdata[xcols*4+i] = int32(-math.Log10(xref.pvalue) * 1000000)
1552         }
1553         return xdata
1554 }
1555
1556 // transpose onehot data from in[col][row] to numpy-style
1557 // out[row*cols+col].
1558 func onehotcols2int8(in [][]int8) []int8 {
1559         if len(in) == 0 {
1560                 return nil
1561         }
1562         cols := len(in)
1563         rows := len(in[0])
1564         out := make([]int8, rows*cols)
1565         for row := 0; row < rows; row++ {
1566                 outrow := out[row*cols:]
1567                 for col, incol := range in {
1568                         outrow[col] = incol[row]
1569                 }
1570         }
1571         return out
1572 }
1573
1574 // Return [2][]uint32{rowIndices, colIndices} indicating which
1575 // elements of matrixT[c][r] have non-zero values.
1576 func onehotChunk2Indirect(matrixT [][]int8) [2][]uint32 {
1577         var nz [2][]uint32
1578         for c, col := range matrixT {
1579                 for r, val := range col {
1580                         if val != 0 {
1581                                 nz[0] = append(nz[0], uint32(r))
1582                                 nz[1] = append(nz[1], uint32(c))
1583                         }
1584                 }
1585         }
1586         return nz
1587 }