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         chi2CaseControlColumn string
  45         chi2CaseControlFile   string
  46         chi2Cases             []bool
  47         chi2PValue            float64
  48         trainingSet           []int // see loadTrainingSet
  49         trainingSetSize       int
  50         minCoverage           int
  51         cgnames               []string
  52         includeVariant1       bool
  53         debugTag              tagID
  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         trainingSetFilename := flags.String("training-set", "", "`tsv` file with sample IDs to be used for PCA fitting and Χ² test (if not provided, use all 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.StringVar(&cmd.chi2CaseControlFile, "chi2-case-control-file", "", "tsv file or directory indicating cases and controls for Χ² test (if directory, all .tsv files will be read)")
  91         flags.StringVar(&cmd.chi2CaseControlColumn, "chi2-case-control-column", "", "name of case/control column in case-control files for Χ² test (value must be 0 for control, 1 for case)")
  92         flags.Float64Var(&cmd.chi2PValue, "chi2-p-value", 1, "do Χ² test and omit columns with p-value above this threshold")
  93         flags.BoolVar(&cmd.includeVariant1, "include-variant-1", false, "include most common variant when building one-hot matrix")
  94         cmd.filter.Flags(flags)
  95         err := flags.Parse(args)
  96         if err == flag.ErrHelp {
  97                 return nil
  98         } else if err != nil {
  99                 return err
 100         }
 101
 102         if *pprof != "" {
 103                 go func() {
 104                         log.Println(http.ListenAndServe(*pprof, nil))
 105                 }()
 106         }
 107
 108         if cmd.chi2PValue != 1 && (cmd.chi2CaseControlFile == "" || cmd.chi2CaseControlColumn == "") {
 109                 return fmt.Errorf("cannot use provided -chi2-p-value=%f because -chi2-case-control-file= or -chi2-case-control-column= value is empty", cmd.chi2PValue)
 110         }
 111
 112         cmd.debugTag = tagID(*debugTag)
 113
 114         if !*runlocal {
 115                 runner := arvadosContainerRunner{
 116                         Name:        "lightning slice-numpy",
 117                         Client:      arvados.NewClientFromEnv(),
 118                         ProjectUUID: *projectUUID,
 119                         RAM:         int64(*arvadosRAM),
 120                         VCPUs:       *arvadosVCPUs,
 121                         Priority:    *priority,
 122                         KeepCache:   2,
 123                         APIAccess:   true,
 124                 }
 125                 err = runner.TranslatePaths(inputDir, regionsFilename, trainingSetFilename, &cmd.chi2CaseControlFile)
 126                 if err != nil {
 127                         return err
 128                 }
 129                 runner.Args = []string{"slice-numpy", "-local=true",
 130                         "-pprof=:6060",
 131                         "-input-dir=" + *inputDir,
 132                         "-output-dir=/mnt/output",
 133                         "-threads=" + fmt.Sprintf("%d", cmd.threads),
 134                         "-regions=" + *regionsFilename,
 135                         "-expand-regions=" + fmt.Sprintf("%d", *expandRegions),
 136                         "-merge-output=" + fmt.Sprintf("%v", *mergeOutput),
 137                         "-single-hgvs-matrix=" + fmt.Sprintf("%v", *hgvsSingle),
 138                         "-chunked-hgvs-matrix=" + fmt.Sprintf("%v", *hgvsChunked),
 139                         "-single-onehot=" + fmt.Sprintf("%v", *onehotSingle),
 140                         "-chunked-onehot=" + fmt.Sprintf("%v", *onehotChunked),
 141                         "-training-set=" + *trainingSetFilename,
 142                         "-pca=" + fmt.Sprintf("%v", *onlyPCA),
 143                         "-pca-components=" + fmt.Sprintf("%d", *pcaComponents),
 144                         "-max-pca-tiles=" + fmt.Sprintf("%d", *maxPCATiles),
 145                         "-chi2-case-control-file=" + cmd.chi2CaseControlFile,
 146                         "-chi2-case-control-column=" + cmd.chi2CaseControlColumn,
 147                         "-chi2-p-value=" + fmt.Sprintf("%f", cmd.chi2PValue),
 148                         "-include-variant-1=" + fmt.Sprintf("%v", cmd.includeVariant1),
 149                         "-debug-tag=" + fmt.Sprintf("%d", cmd.debugTag),
 150                 }
 151                 runner.Args = append(runner.Args, cmd.filter.Args()...)
 152                 var output string
 153                 output, err = runner.Run()
 154                 if err != nil {
 155                         return err
 156                 }
 157                 fmt.Fprintln(stdout, output)
 158                 return nil
 159         }
 160
 161         infiles, err := allFiles(*inputDir, matchGobFile)
 162         if err != nil {
 163                 return err
 164         }
 165         if len(infiles) == 0 {
 166                 err = fmt.Errorf("no input files found in %s", *inputDir)
 167                 return err
 168         }
 169         sort.Strings(infiles)
 170
 171         var refseq map[string][]tileLibRef
 172         var reftiledata = make(map[tileLibRef][]byte, 11000000)
 173         in0, err := open(infiles[0])
 174         if err != nil {
 175                 return err
 176         }
 177
 178         matchGenome, err := regexp.Compile(cmd.filter.MatchGenome)
 179         if err != nil {
 180                 err = fmt.Errorf("-match-genome: invalid regexp: %q", cmd.filter.MatchGenome)
 181                 return err
 182         }
 183
 184         cmd.cgnames = nil
 185         var tagset [][]byte
 186         err = DecodeLibrary(in0, strings.HasSuffix(infiles[0], ".gz"), func(ent *LibraryEntry) error {
 187                 if len(ent.TagSet) > 0 {
 188                         tagset = ent.TagSet
 189                 }
 190                 for _, cseq := range ent.CompactSequences {
 191                         if cseq.Name == *ref || *ref == "" {
 192                                 refseq = cseq.TileSequences
 193                         }
 194                 }
 195                 for _, cg := range ent.CompactGenomes {
 196                         if matchGenome.MatchString(cg.Name) {
 197                                 cmd.cgnames = append(cmd.cgnames, cg.Name)
 198                         }
 199                 }
 200                 for _, tv := range ent.TileVariants {
 201                         if tv.Ref {
 202                                 reftiledata[tileLibRef{tv.Tag, tv.Variant}] = tv.Sequence
 203                         }
 204                 }
 205                 return nil
 206         })
 207         if err != nil {
 208                 return err
 209         }
 210         in0.Close()
 211         if refseq == nil {
 212                 err = fmt.Errorf("%s: reference sequence not found", infiles[0])
 213                 return err
 214         }
 215         if len(tagset) == 0 {
 216                 err = fmt.Errorf("tagset not found")
 217                 return err
 218         }
 219
 220         taglib := &tagLibrary{}
 221         err = taglib.setTags(tagset)
 222         if err != nil {
 223                 return err
 224         }
 225         taglen := taglib.TagLen()
 226
 227         if len(cmd.cgnames) == 0 {
 228                 err = fmt.Errorf("no genomes found matching regexp %q", cmd.filter.MatchGenome)
 229                 return err
 230         }
 231         sort.Strings(cmd.cgnames)
 232         err = cmd.useCaseControlFiles()
 233         if err != nil {
 234                 return err
 235         }
 236         if len(cmd.cgnames) == 0 {
 237                 err = fmt.Errorf("fatal: 0 cases, 0 controls, nothing to do")
 238                 return err
 239         }
 240         err = cmd.loadTrainingSet(*trainingSetFilename)
 241         if err != nil {
 242                 return err
 243         }
 244         if cmd.filter.MinCoverage == 1 {
 245                 // In the generic formula below, floating point
 246                 // arithmetic can effectively push the coverage
 247                 // threshold above 1.0, which is impossible/useless.
 248                 // 1.0 needs to mean exactly 100% coverage.
 249                 cmd.minCoverage = len(cmd.cgnames)
 250         } else {
 251                 cmd.minCoverage = int(math.Ceil(cmd.filter.MinCoverage * float64(len(cmd.cgnames))))
 252         }
 253
 254         {
 255                 labelsFilename := *outputDir + "/samples.csv"
 256                 log.Infof("writing labels to %s", labelsFilename)
 257                 var f *os.File
 258                 f, err = os.Create(labelsFilename)
 259                 if err != nil {
 260                         return err
 261                 }
 262                 defer f.Close()
 263                 for i, name := range cmd.cgnames {
 264                         cc := 0
 265                         if cmd.chi2Cases != nil && cmd.chi2Cases[i] {
 266                                 cc = 1
 267                         }
 268                         _, err = fmt.Fprintf(f, "%d,%q,%d\n", i, trimFilenameForLabel(name), cc)
 269                         if err != nil {
 270                                 err = fmt.Errorf("write %s: %w", labelsFilename, err)
 271                                 return err
 272                         }
 273                 }
 274                 err = f.Close()
 275                 if err != nil {
 276                         err = fmt.Errorf("close %s: %w", labelsFilename, err)
 277                         return err
 278                 }
 279         }
 280
 281         log.Info("indexing reference tiles")
 282         type reftileinfo struct {
 283                 variant  tileVariantID
 284                 seqname  string // chr1
 285                 pos      int    // distance from start of chromosome to starttag
 286                 tiledata []byte // acgtggcaa...
 287                 excluded bool   // true if excluded by regions file
 288                 nexttag  tagID  // tagID of following tile (-1 for last tag of chromosome)
 289         }
 290         isdup := map[tagID]bool{}
 291         reftile := map[tagID]*reftileinfo{}
 292         for seqname, cseq := range refseq {
 293                 pos := 0
 294                 lastreftag := tagID(-1)
 295                 for _, libref := range cseq {
 296                         if cmd.filter.MaxTag >= 0 && libref.Tag > tagID(cmd.filter.MaxTag) {
 297                                 continue
 298                         }
 299                         tiledata := reftiledata[libref]
 300                         if len(tiledata) == 0 {
 301                                 err = fmt.Errorf("missing tiledata for tag %d variant %d in %s in ref", libref.Tag, libref.Variant, seqname)
 302                                 return err
 303                         }
 304                         foundthistag := false
 305                         taglib.FindAll(tiledata[:len(tiledata)-1], func(tagid tagID, offset, _ int) {
 306                                 if !foundthistag && tagid == libref.Tag {
 307                                         foundthistag = true
 308                                         return
 309                                 }
 310                                 if dupref, ok := reftile[tagid]; ok {
 311                                         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)
 312                                         delete(reftile, tagid)
 313                                 } else {
 314                                         log.Printf("found tag %d at offset %d inside tile variant %+v on %s @ %d", tagid, offset, libref, seqname, pos+offset+1)
 315                                 }
 316                                 isdup[tagid] = true
 317                         })
 318                         if isdup[libref.Tag] {
 319                                 log.Printf("dropping reference tile %+v from %s @ %d, tag not unique", libref, seqname, pos)
 320                         } else if reftile[libref.Tag] != nil {
 321                                 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)
 322                                 delete(reftile, libref.Tag)
 323                                 log.Printf("dropping reference tile %+v from %s @ %d, tag not unique", libref, seqname, pos)
 324                                 isdup[libref.Tag] = true
 325                         } else {
 326                                 reftile[libref.Tag] = &reftileinfo{
 327                                         seqname:  seqname,
 328                                         variant:  libref.Variant,
 329                                         tiledata: tiledata,
 330                                         pos:      pos,
 331                                         nexttag:  -1,
 332                                 }
 333                                 if lastreftag >= 0 {
 334                                         reftile[lastreftag].nexttag = libref.Tag
 335                                 }
 336                                 lastreftag = libref.Tag
 337                         }
 338                         pos += len(tiledata) - taglen
 339                 }
 340                 log.Printf("... %s done, len %d", seqname, pos+taglen)
 341         }
 342
 343         var mask *mask
 344         if *regionsFilename != "" {
 345                 log.Printf("loading regions from %s", *regionsFilename)
 346                 mask, err = makeMask(*regionsFilename, *expandRegions)
 347                 if err != nil {
 348                         return err
 349                 }
 350                 log.Printf("before applying mask, len(reftile) == %d", len(reftile))
 351                 log.Printf("deleting reftile entries for regions outside %d intervals", mask.Len())
 352                 for _, rt := range reftile {
 353                         if !mask.Check(strings.TrimPrefix(rt.seqname, "chr"), rt.pos, rt.pos+len(rt.tiledata)) {
 354                                 rt.excluded = true
 355                         }
 356                 }
 357                 log.Printf("after applying mask, len(reftile) == %d", len(reftile))
 358         }
 359
 360         type hgvsColSet map[hgvs.Variant][2][]int8
 361         encodeHGVS := throttle{Max: len(refseq)}
 362         encodeHGVSTodo := map[string]chan hgvsColSet{}
 363         tmpHGVSCols := map[string]*os.File{}
 364         if *hgvsChunked {
 365                 for seqname := range refseq {
 366                         var f *os.File
 367                         f, err = os.Create(*outputDir + "/tmp." + seqname + ".gob")
 368                         if err != nil {
 369                                 return err
 370                         }
 371                         defer os.Remove(f.Name())
 372                         bufw := bufio.NewWriterSize(f, 1<<24)
 373                         enc := gob.NewEncoder(bufw)
 374                         tmpHGVSCols[seqname] = f
 375                         todo := make(chan hgvsColSet, 128)
 376                         encodeHGVSTodo[seqname] = todo
 377                         encodeHGVS.Go(func() error {
 378                                 for colset := range todo {
 379                                         err := enc.Encode(colset)
 380                                         if err != nil {
 381                                                 encodeHGVS.Report(err)
 382                                                 for range todo {
 383                                                 }
 384                                                 return err
 385                                         }
 386                                 }
 387                                 return bufw.Flush()
 388                         })
 389                 }
 390         }
 391
 392         var toMerge [][]int16
 393         if *mergeOutput || *hgvsSingle {
 394                 toMerge = make([][]int16, len(infiles))
 395         }
 396         var onehotIndirect [][2][]uint32 // [chunkIndex][axis][index]
 397         var onehotChunkSize []uint32
 398         var onehotXrefs [][]onehotXref
 399         if *onehotSingle || *onlyPCA {
 400                 onehotIndirect = make([][2][]uint32, len(infiles))
 401                 onehotChunkSize = make([]uint32, len(infiles))
 402                 onehotXrefs = make([][]onehotXref, len(infiles))
 403         }
 404         chunkStartTag := make([]tagID, len(infiles))
 405
 406         throttleMem := throttle{Max: cmd.threads} // TODO: estimate using mem and data size
 407         throttleNumpyMem := throttle{Max: cmd.threads/2 + 1}
 408         log.Info("generating annotations and numpy matrix for each slice")
 409         var errSkip = errors.New("skip infile")
 410         var done int64
 411         for infileIdx, infile := range infiles {
 412                 infileIdx, infile := infileIdx, infile
 413                 throttleMem.Go(func() error {
 414                         seq := make(map[tagID][]TileVariant, 50000)
 415                         cgs := make(map[string]CompactGenome, len(cmd.cgnames))
 416                         f, err := open(infile)
 417                         if err != nil {
 418                                 return err
 419                         }
 420                         defer f.Close()
 421                         log.Infof("%04d: reading %s", infileIdx, infile)
 422                         err = DecodeLibrary(f, strings.HasSuffix(infile, ".gz"), func(ent *LibraryEntry) error {
 423                                 for _, tv := range ent.TileVariants {
 424                                         if tv.Ref {
 425                                                 continue
 426                                         }
 427                                         // Skip tile with no
 428                                         // corresponding ref tile, if
 429                                         // mask is in play (we can't
 430                                         // determine coordinates for
 431                                         // these)
 432                                         if mask != nil && reftile[tv.Tag] == nil {
 433                                                 continue
 434                                         }
 435                                         // Skip tile whose
 436                                         // corresponding ref tile is
 437                                         // outside target regions --
 438                                         // unless it's a potential
 439                                         // spanning tile.
 440                                         if mask != nil && reftile[tv.Tag].excluded &&
 441                                                 (int(tv.Tag+1) >= len(tagset) ||
 442                                                         (bytes.HasSuffix(tv.Sequence, tagset[tv.Tag+1]) && reftile[tv.Tag+1] != nil && !reftile[tv.Tag+1].excluded)) {
 443                                                 continue
 444                                         }
 445                                         if tv.Tag == cmd.debugTag {
 446                                                 log.Printf("infile %d %s tag %d variant %d hash %x", infileIdx, infile, tv.Tag, tv.Variant, tv.Blake2b[:3])
 447                                         }
 448                                         variants := seq[tv.Tag]
 449                                         if len(variants) == 0 {
 450                                                 variants = make([]TileVariant, 100)
 451                                         }
 452                                         for len(variants) <= int(tv.Variant) {
 453                                                 variants = append(variants, TileVariant{})
 454                                         }
 455                                         variants[int(tv.Variant)] = tv
 456                                         seq[tv.Tag] = variants
 457                                 }
 458                                 for _, cg := range ent.CompactGenomes {
 459                                         if cmd.filter.MaxTag >= 0 && cg.StartTag > tagID(cmd.filter.MaxTag) {
 460                                                 return errSkip
 461                                         }
 462                                         if !matchGenome.MatchString(cg.Name) {
 463                                                 continue
 464                                         }
 465                                         // pad to full slice size
 466                                         // to avoid out-of-bounds
 467                                         // checks later
 468                                         if sliceSize := 2 * int(cg.EndTag-cg.StartTag); len(cg.Variants) < sliceSize {
 469                                                 cg.Variants = append(cg.Variants, make([]tileVariantID, sliceSize-len(cg.Variants))...)
 470                                         }
 471                                         cgs[cg.Name] = cg
 472                                 }
 473                                 return nil
 474                         })
 475                         if err == errSkip {
 476                                 return nil
 477                         } else if err != nil {
 478                                 return fmt.Errorf("%04d: DecodeLibrary(%s): err", infileIdx, infile)
 479                         }
 480                         tagstart := cgs[cmd.cgnames[0]].StartTag
 481                         tagend := cgs[cmd.cgnames[0]].EndTag
 482                         chunkStartTag[infileIdx] = tagstart
 483
 484                         // TODO: filters
 485
 486                         log.Infof("%04d: renumber/dedup variants for tags %d-%d", infileIdx, tagstart, tagend)
 487                         variantRemap := make([][]tileVariantID, tagend-tagstart)
 488                         throttleCPU := throttle{Max: runtime.GOMAXPROCS(0)}
 489                         for tag, variants := range seq {
 490                                 tag, variants := tag, variants
 491                                 throttleCPU.Go(func() error {
 492                                         alleleCoverage := 0
 493                                         count := make(map[[blake2b.Size256]byte]int, len(variants))
 494
 495                                         rt := reftile[tag]
 496                                         if rt != nil {
 497                                                 count[blake2b.Sum256(rt.tiledata)] = 0
 498                                         }
 499
 500                                         for cgname, cg := range cgs {
 501                                                 idx := int(tag-tagstart) * 2
 502                                                 for allele := 0; allele < 2; allele++ {
 503                                                         v := cg.Variants[idx+allele]
 504                                                         if v > 0 && len(variants[v].Sequence) > 0 {
 505                                                                 count[variants[v].Blake2b]++
 506                                                                 alleleCoverage++
 507                                                         }
 508                                                         if v > 0 && tag == cmd.debugTag {
 509                                                                 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])
 510                                                         }
 511                                                 }
 512                                         }
 513                                         if alleleCoverage < cmd.minCoverage*2 {
 514                                                 idx := int(tag-tagstart) * 2
 515                                                 for _, cg := range cgs {
 516                                                         cg.Variants[idx] = 0
 517                                                         cg.Variants[idx+1] = 0
 518                                                 }
 519                                                 if tag == cmd.debugTag {
 520                                                         log.Printf("tag %d alleleCoverage %d < min %d, sample data wiped", tag, alleleCoverage, cmd.minCoverage*2)
 521                                                 }
 522                                                 return nil
 523                                         }
 524
 525                                         // hash[i] will be the hash of
 526                                         // the variant(s) that should
 527                                         // be at rank i (0-based).
 528                                         hash := make([][blake2b.Size256]byte, 0, len(count))
 529                                         for b := range count {
 530                                                 hash = append(hash, b)
 531                                         }
 532                                         sort.Slice(hash, func(i, j int) bool {
 533                                                 bi, bj := &hash[i], &hash[j]
 534                                                 if ci, cj := count[*bi], count[*bj]; ci != cj {
 535                                                         return ci > cj
 536                                                 } else {
 537                                                         return bytes.Compare((*bi)[:], (*bj)[:]) < 0
 538                                                 }
 539                                         })
 540                                         // rank[b] will be the 1-based
 541                                         // new variant number for
 542                                         // variants whose hash is b.
 543                                         rank := make(map[[blake2b.Size256]byte]tileVariantID, len(hash))
 544                                         for i, h := range hash {
 545                                                 rank[h] = tileVariantID(i + 1)
 546                                         }
 547                                         if tag == cmd.debugTag {
 548                                                 for h, r := range rank {
 549                                                         log.Printf("tag %d rank(%x) = %v", tag, h[:3], r)
 550                                                 }
 551                                         }
 552                                         // remap[v] will be the new
 553                                         // variant number for original
 554                                         // variant number v.
 555                                         remap := make([]tileVariantID, len(variants))
 556                                         for i, tv := range variants {
 557                                                 remap[i] = rank[tv.Blake2b]
 558                                         }
 559                                         if tag == cmd.debugTag {
 560                                                 for in, out := range remap {
 561                                                         if out > 0 {
 562                                                                 log.Printf("tag %d remap %d => %d", tag, in, out)
 563                                                         }
 564                                                 }
 565                                         }
 566                                         variantRemap[tag-tagstart] = remap
 567                                         if rt != nil {
 568                                                 refrank := rank[blake2b.Sum256(rt.tiledata)]
 569                                                 if tag == cmd.debugTag {
 570                                                         log.Printf("tag %d reftile variant %d => %d", tag, rt.variant, refrank)
 571                                                 }
 572                                                 rt.variant = refrank
 573                                         }
 574                                         return nil
 575                                 })
 576                         }
 577                         throttleCPU.Wait()
 578
 579                         var onehotChunk [][]int8
 580                         var onehotXref []onehotXref
 581
 582                         var annotationsFilename string
 583                         if *onlyPCA {
 584                                 annotationsFilename = "/dev/null"
 585                         } else {
 586                                 annotationsFilename = fmt.Sprintf("%s/matrix.%04d.annotations.csv", *outputDir, infileIdx)
 587                                 log.Infof("%04d: writing %s", infileIdx, annotationsFilename)
 588                         }
 589                         annof, err := os.Create(annotationsFilename)
 590                         if err != nil {
 591                                 return err
 592                         }
 593                         annow := bufio.NewWriterSize(annof, 1<<20)
 594                         outcol := 0
 595                         for tag := tagstart; tag < tagend; tag++ {
 596                                 rt := reftile[tag]
 597                                 if rt == nil && mask != nil {
 598                                         // With no ref tile, we don't
 599                                         // have coordinates to say
 600                                         // this is in the desired
 601                                         // regions -- so it's not.
 602                                         // TODO: handle ref spanning
 603                                         // tile case.
 604                                         continue
 605                                 }
 606                                 if rt != nil && rt.excluded {
 607                                         // TODO: don't skip yet --
 608                                         // first check for spanning
 609                                         // tile variants that
 610                                         // intersect non-excluded ref
 611                                         // tiles.
 612                                         continue
 613                                 }
 614                                 if cmd.filter.MaxTag >= 0 && tag > tagID(cmd.filter.MaxTag) {
 615                                         break
 616                                 }
 617                                 remap := variantRemap[tag-tagstart]
 618                                 maxv := tileVariantID(0)
 619                                 for _, v := range remap {
 620                                         if maxv < v {
 621                                                 maxv = v
 622                                         }
 623                                 }
 624                                 if *onehotChunked || *onehotSingle || *onlyPCA {
 625                                         onehot, xrefs := cmd.tv2homhet(cgs, maxv, remap, tag, tagstart, seq)
 626                                         if tag == cmd.debugTag {
 627                                                 log.WithFields(logrus.Fields{
 628                                                         "onehot": onehot,
 629                                                         "xrefs":  xrefs,
 630                                                 }).Info("tv2homhet()")
 631                                         }
 632                                         onehotChunk = append(onehotChunk, onehot...)
 633                                         onehotXref = append(onehotXref, xrefs...)
 634                                 }
 635                                 if *onlyPCA {
 636                                         outcol++
 637                                         continue
 638                                 }
 639                                 if rt == nil {
 640                                         // Reference does not use any
 641                                         // variant of this tile
 642                                         //
 643                                         // TODO: diff against the
 644                                         // relevant portion of the
 645                                         // ref's spanning tile
 646                                         outcol++
 647                                         continue
 648                                 }
 649                                 fmt.Fprintf(annow, "%d,%d,%d,=,%s,%d,,,\n", tag, outcol, rt.variant, rt.seqname, rt.pos)
 650                                 variants := seq[tag]
 651                                 reftilestr := strings.ToUpper(string(rt.tiledata))
 652
 653                                 done := make([]bool, maxv+1)
 654                                 variantDiffs := make([][]hgvs.Variant, maxv+1)
 655                                 for v, tv := range variants {
 656                                         v := remap[v]
 657                                         if v == 0 || v == rt.variant || done[v] {
 658                                                 continue
 659                                         } else {
 660                                                 done[v] = true
 661                                         }
 662                                         if len(tv.Sequence) < taglen {
 663                                                 continue
 664                                         }
 665                                         // if reftilestr doesn't end
 666                                         // in the same tag as tv,
 667                                         // extend reftilestr with
 668                                         // following ref tiles until
 669                                         // it does (up to an arbitrary
 670                                         // sanity-check limit)
 671                                         reftilestr := reftilestr
 672                                         endtagstr := strings.ToUpper(string(tv.Sequence[len(tv.Sequence)-taglen:]))
 673                                         for i, rt := 0, rt; i < annotationMaxTileSpan && !strings.HasSuffix(reftilestr, endtagstr) && rt.nexttag >= 0; i++ {
 674                                                 rt = reftile[rt.nexttag]
 675                                                 if rt == nil {
 676                                                         break
 677                                                 }
 678                                                 reftilestr += strings.ToUpper(string(rt.tiledata[taglen:]))
 679                                         }
 680                                         if mask != nil && !mask.Check(strings.TrimPrefix(rt.seqname, "chr"), rt.pos, rt.pos+len(reftilestr)) {
 681                                                 continue
 682                                         }
 683                                         if !strings.HasSuffix(reftilestr, endtagstr) {
 684                                                 fmt.Fprintf(annow, "%d,%d,%d,,%s,%d,,,\n", tag, outcol, v, rt.seqname, rt.pos)
 685                                                 continue
 686                                         }
 687                                         if lendiff := len(reftilestr) - len(tv.Sequence); lendiff < -1000 || lendiff > 1000 {
 688                                                 fmt.Fprintf(annow, "%d,%d,%d,,%s,%d,,,\n", tag, outcol, v, rt.seqname, rt.pos)
 689                                                 continue
 690                                         }
 691                                         diffs, _ := hgvs.Diff(reftilestr, strings.ToUpper(string(tv.Sequence)), 0)
 692                                         for i := range diffs {
 693                                                 diffs[i].Position += rt.pos
 694                                         }
 695                                         for _, diff := range diffs {
 696                                                 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)
 697                                         }
 698                                         if *hgvsChunked {
 699                                                 variantDiffs[v] = diffs
 700                                         }
 701                                 }
 702                                 if *hgvsChunked {
 703                                         // We can now determine, for each HGVS
 704                                         // variant (diff) in this reftile
 705                                         // region, whether a given genome
 706                                         // phase/allele (1) has the variant, (0) has
 707                                         // =ref or a different variant in that
 708                                         // position, or (-1) is lacking
 709                                         // coverage / couldn't be diffed.
 710                                         hgvsCol := hgvsColSet{}
 711                                         for _, diffs := range variantDiffs {
 712                                                 for _, diff := range diffs {
 713                                                         if _, ok := hgvsCol[diff]; ok {
 714                                                                 continue
 715                                                         }
 716                                                         hgvsCol[diff] = [2][]int8{
 717                                                                 make([]int8, len(cmd.cgnames)),
 718                                                                 make([]int8, len(cmd.cgnames)),
 719                                                         }
 720                                                 }
 721                                         }
 722                                         for row, name := range cmd.cgnames {
 723                                                 variants := cgs[name].Variants[(tag-tagstart)*2:]
 724                                                 for ph := 0; ph < 2; ph++ {
 725                                                         v := variants[ph]
 726                                                         if int(v) >= len(remap) {
 727                                                                 v = 0
 728                                                         } else {
 729                                                                 v = remap[v]
 730                                                         }
 731                                                         if v == rt.variant {
 732                                                                 // hgvsCol[*][ph][row] is already 0
 733                                                         } else if len(variantDiffs[v]) == 0 {
 734                                                                 // lacking coverage / couldn't be diffed
 735                                                                 for _, col := range hgvsCol {
 736                                                                         col[ph][row] = -1
 737                                                                 }
 738                                                         } else {
 739                                                                 for _, diff := range variantDiffs[v] {
 740                                                                         hgvsCol[diff][ph][row] = 1
 741                                                                 }
 742                                                         }
 743                                                 }
 744                                         }
 745                                         for diff, colpair := range hgvsCol {
 746                                                 allele2homhet(colpair)
 747                                                 if !cmd.filterHGVScolpair(colpair) {
 748                                                         delete(hgvsCol, diff)
 749                                                 }
 750                                         }
 751                                         if len(hgvsCol) > 0 {
 752                                                 encodeHGVSTodo[rt.seqname] <- hgvsCol
 753                                         }
 754                                 }
 755                                 outcol++
 756                         }
 757                         err = annow.Flush()
 758                         if err != nil {
 759                                 return err
 760                         }
 761                         err = annof.Close()
 762                         if err != nil {
 763                                 return err
 764                         }
 765
 766                         if *onehotChunked {
 767                                 // transpose onehotChunk[col][row] to numpy[row*ncols+col]
 768                                 rows := len(cmd.cgnames)
 769                                 cols := len(onehotChunk)
 770                                 log.Infof("%04d: preparing onehot numpy (rows=%d, cols=%d, mem=%d)", infileIdx, rows, cols, rows*cols)
 771                                 throttleNumpyMem.Acquire()
 772                                 out := onehotcols2int8(onehotChunk)
 773                                 fnm := fmt.Sprintf("%s/onehot.%04d.npy", *outputDir, infileIdx)
 774                                 err = writeNumpyInt8(fnm, out, rows, cols)
 775                                 if err != nil {
 776                                         return err
 777                                 }
 778                                 fnm = fmt.Sprintf("%s/onehot-columns.%04d.npy", *outputDir, infileIdx)
 779                                 err = writeNumpyInt32(fnm, onehotXref2int32(onehotXref), 4, len(onehotXref))
 780                                 if err != nil {
 781                                         return err
 782                                 }
 783                                 debug.FreeOSMemory()
 784                                 throttleNumpyMem.Release()
 785                         }
 786                         if *onehotSingle || *onlyPCA {
 787                                 onehotIndirect[infileIdx] = onehotChunk2Indirect(onehotChunk)
 788                                 onehotChunkSize[infileIdx] = uint32(len(onehotChunk))
 789                                 onehotXrefs[infileIdx] = onehotXref
 790                                 n := len(onehotIndirect[infileIdx][0])
 791                                 log.Infof("%04d: keeping onehot coordinates in memory (n=%d, mem=%d)", infileIdx, n, n*8*2)
 792                         }
 793                         if !(*onehotSingle || *onehotChunked || *onlyPCA) || *mergeOutput || *hgvsSingle {
 794                                 log.Infof("%04d: preparing numpy (rows=%d, cols=%d)", infileIdx, len(cmd.cgnames), 2*outcol)
 795                                 throttleNumpyMem.Acquire()
 796                                 rows := len(cmd.cgnames)
 797                                 cols := 2 * outcol
 798                                 out := make([]int16, rows*cols)
 799                                 for row, name := range cmd.cgnames {
 800                                         outidx := row * cols
 801                                         for col, v := range cgs[name].Variants {
 802                                                 tag := tagstart + tagID(col/2)
 803                                                 if cmd.filter.MaxTag >= 0 && tag > tagID(cmd.filter.MaxTag) {
 804                                                         break
 805                                                 }
 806                                                 if rt := reftile[tag]; rt == nil || rt.excluded {
 807                                                         continue
 808                                                 }
 809                                                 if v == 0 {
 810                                                         out[outidx] = 0 // tag not found / spanning tile
 811                                                 } else if variants, ok := seq[tag]; ok && int(v) < len(variants) && len(variants[v].Sequence) > 0 {
 812                                                         out[outidx] = int16(variantRemap[tag-tagstart][v])
 813                                                 } else {
 814                                                         out[outidx] = -1 // low quality tile variant
 815                                                 }
 816                                                 if tag == cmd.debugTag {
 817                                                         log.Printf("tag %d row %d col %d outidx %d v %d out %d", tag, row, col, outidx, v, out[outidx])
 818                                                 }
 819                                                 outidx++
 820                                         }
 821                                 }
 822                                 seq = nil
 823                                 cgs = nil
 824                                 debug.FreeOSMemory()
 825                                 throttleNumpyMem.Release()
 826                                 if *mergeOutput || *hgvsSingle {
 827                                         log.Infof("%04d: matrix fragment %d rows x %d cols", infileIdx, rows, cols)
 828                                         toMerge[infileIdx] = out
 829                                 }
 830                                 if !*mergeOutput && !*onehotChunked && !*onehotSingle {
 831                                         fnm := fmt.Sprintf("%s/matrix.%04d.npy", *outputDir, infileIdx)
 832                                         err = writeNumpyInt16(fnm, out, rows, cols)
 833                                         if err != nil {
 834                                                 return err
 835                                         }
 836                                 }
 837                         }
 838                         debug.FreeOSMemory()
 839                         log.Infof("%s: done (%d/%d)", infile, int(atomic.AddInt64(&done, 1)), len(infiles))
 840                         return nil
 841                 })
 842         }
 843         if err = throttleMem.Wait(); err != nil {
 844                 return err
 845         }
 846
 847         if *hgvsChunked {
 848                 log.Info("flushing hgvsCols temp files")
 849                 for seqname := range refseq {
 850                         close(encodeHGVSTodo[seqname])
 851                 }
 852                 err = encodeHGVS.Wait()
 853                 if err != nil {
 854                         return err
 855                 }
 856                 for seqname := range refseq {
 857                         log.Infof("%s: reading hgvsCols from temp file", seqname)
 858                         f := tmpHGVSCols[seqname]
 859                         _, err = f.Seek(0, io.SeekStart)
 860                         if err != nil {
 861                                 return err
 862                         }
 863                         var hgvsCols hgvsColSet
 864                         dec := gob.NewDecoder(bufio.NewReaderSize(f, 1<<24))
 865                         for err == nil {
 866                                 err = dec.Decode(&hgvsCols)
 867                         }
 868                         if err != io.EOF {
 869                                 return err
 870                         }
 871                         log.Infof("%s: sorting %d hgvs variants", seqname, len(hgvsCols))
 872                         variants := make([]hgvs.Variant, 0, len(hgvsCols))
 873                         for v := range hgvsCols {
 874                                 variants = append(variants, v)
 875                         }
 876                         sort.Slice(variants, func(i, j int) bool {
 877                                 vi, vj := &variants[i], &variants[j]
 878                                 if vi.Position != vj.Position {
 879                                         return vi.Position < vj.Position
 880                                 } else if vi.Ref != vj.Ref {
 881                                         return vi.Ref < vj.Ref
 882                                 } else {
 883                                         return vi.New < vj.New
 884                                 }
 885                         })
 886                         rows := len(cmd.cgnames)
 887                         cols := len(variants) * 2
 888                         log.Infof("%s: building hgvs matrix (rows=%d, cols=%d, mem=%d)", seqname, rows, cols, rows*cols)
 889                         out := make([]int8, rows*cols)
 890                         for varIdx, variant := range variants {
 891                                 hgvsCols := hgvsCols[variant]
 892                                 for row := range cmd.cgnames {
 893                                         for ph := 0; ph < 2; ph++ {
 894                                                 out[row*cols+varIdx+ph] = hgvsCols[ph][row]
 895                                         }
 896                                 }
 897                         }
 898                         err = writeNumpyInt8(fmt.Sprintf("%s/hgvs.%s.npy", *outputDir, seqname), out, rows, cols)
 899                         if err != nil {
 900                                 return err
 901                         }
 902                         out = nil
 903
 904                         fnm := fmt.Sprintf("%s/hgvs.%s.annotations.csv", *outputDir, seqname)
 905                         log.Infof("%s: writing hgvs column labels to %s", seqname, fnm)
 906                         var hgvsLabels bytes.Buffer
 907                         for varIdx, variant := range variants {
 908                                 fmt.Fprintf(&hgvsLabels, "%d,%s:g.%s\n", varIdx, seqname, variant.String())
 909                         }
 910                         err = ioutil.WriteFile(fnm, hgvsLabels.Bytes(), 0666)
 911                         if err != nil {
 912                                 return err
 913                         }
 914                 }
 915         }
 916
 917         if *mergeOutput || *hgvsSingle {
 918                 var annow *bufio.Writer
 919                 var annof *os.File
 920                 if *mergeOutput {
 921                         annoFilename := fmt.Sprintf("%s/matrix.annotations.csv", *outputDir)
 922                         annof, err = os.Create(annoFilename)
 923                         if err != nil {
 924                                 return err
 925                         }
 926                         annow = bufio.NewWriterSize(annof, 1<<20)
 927                 }
 928
 929                 rows := len(cmd.cgnames)
 930                 cols := 0
 931                 for _, chunk := range toMerge {
 932                         cols += len(chunk) / rows
 933                 }
 934                 log.Infof("merging output matrix (rows=%d, cols=%d, mem=%d) and annotations", rows, cols, rows*cols*2)
 935                 var out []int16
 936                 if *mergeOutput {
 937                         out = make([]int16, rows*cols)
 938                 }
 939                 hgvsCols := map[string][2][]int16{} // hgvs -> [[g0,g1,g2,...], [g0,g1,g2,...]] (slice of genomes for each phase)
 940                 startcol := 0
 941                 for outIdx, chunk := range toMerge {
 942                         chunkcols := len(chunk) / rows
 943                         if *mergeOutput {
 944                                 for row := 0; row < rows; row++ {
 945                                         copy(out[row*cols+startcol:], chunk[row*chunkcols:(row+1)*chunkcols])
 946                                 }
 947                         }
 948                         toMerge[outIdx] = nil
 949
 950                         annotationsFilename := fmt.Sprintf("%s/matrix.%04d.annotations.csv", *outputDir, outIdx)
 951                         log.Infof("reading %s", annotationsFilename)
 952                         buf, err := os.ReadFile(annotationsFilename)
 953                         if err != nil {
 954                                 return err
 955                         }
 956                         if *mergeOutput {
 957                                 err = os.Remove(annotationsFilename)
 958                                 if err != nil {
 959                                         return err
 960                                 }
 961                         }
 962                         for _, line := range bytes.Split(buf, []byte{'\n'}) {
 963                                 if len(line) == 0 {
 964                                         continue
 965                                 }
 966                                 fields := bytes.SplitN(line, []byte{','}, 9)
 967                                 tag, _ := strconv.Atoi(string(fields[0]))
 968                                 incol, _ := strconv.Atoi(string(fields[1]))
 969                                 tileVariant, _ := strconv.Atoi(string(fields[2]))
 970                                 hgvsID := string(fields[3])
 971                                 seqname := string(fields[4])
 972                                 pos, _ := strconv.Atoi(string(fields[5]))
 973                                 refseq := fields[6]
 974                                 if hgvsID == "" {
 975                                         // Null entry for un-diffable
 976                                         // tile variant
 977                                         continue
 978                                 }
 979                                 if hgvsID == "=" {
 980                                         // Null entry for ref tile
 981                                         continue
 982                                 }
 983                                 if mask != nil && !mask.Check(strings.TrimPrefix(seqname, "chr"), pos, pos+len(refseq)) {
 984                                         // The tile intersects one of
 985                                         // the selected regions, but
 986                                         // this particular HGVS
 987                                         // variant does not.
 988                                         continue
 989                                 }
 990                                 hgvsColPair := hgvsCols[hgvsID]
 991                                 if hgvsColPair[0] == nil {
 992                                         // values in new columns start
 993                                         // out as -1 ("no data yet")
 994                                         // or 0 ("=ref") here, may
 995                                         // change to 1 ("hgvs variant
 996                                         // present") below, either on
 997                                         // this line or a future line.
 998                                         hgvsColPair = [2][]int16{make([]int16, len(cmd.cgnames)), make([]int16, len(cmd.cgnames))}
 999                                         rt, ok := reftile[tagID(tag)]
1000                                         if !ok {
1001                                                 err = fmt.Errorf("bug: seeing annotations for tag %d, but it has no reftile entry", tag)
1002                                                 return err
1003                                         }
1004                                         for ph := 0; ph < 2; ph++ {
1005                                                 for row := 0; row < rows; row++ {
1006                                                         v := chunk[row*chunkcols+incol*2+ph]
1007                                                         if tileVariantID(v) == rt.variant {
1008                                                                 hgvsColPair[ph][row] = 0
1009                                                         } else {
1010                                                                 hgvsColPair[ph][row] = -1
1011                                                         }
1012                                                 }
1013                                         }
1014                                         hgvsCols[hgvsID] = hgvsColPair
1015                                         if annow != nil {
1016                                                 hgvsref := hgvs.Variant{
1017                                                         Position: pos,
1018                                                         Ref:      string(refseq),
1019                                                         New:      string(refseq),
1020                                                 }
1021                                                 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])
1022                                         }
1023                                 }
1024                                 if annow != nil {
1025                                         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])
1026                                 }
1027                                 for ph := 0; ph < 2; ph++ {
1028                                         for row := 0; row < rows; row++ {
1029                                                 v := chunk[row*chunkcols+incol*2+ph]
1030                                                 if int(v) == tileVariant {
1031                                                         hgvsColPair[ph][row] = 1
1032                                                 }
1033                                         }
1034                                 }
1035                         }
1036
1037                         startcol += chunkcols
1038                 }
1039                 if *mergeOutput {
1040                         err = annow.Flush()
1041                         if err != nil {
1042                                 return err
1043                         }
1044                         err = annof.Close()
1045                         if err != nil {
1046                                 return err
1047                         }
1048                         err = writeNumpyInt16(fmt.Sprintf("%s/matrix.npy", *outputDir), out, rows, cols)
1049                         if err != nil {
1050                                 return err
1051                         }
1052                 }
1053                 out = nil
1054
1055                 if *hgvsSingle {
1056                         cols = len(hgvsCols) * 2
1057                         log.Printf("building hgvs-based matrix: %d rows x %d cols", rows, cols)
1058                         out = make([]int16, rows*cols)
1059                         hgvsIDs := make([]string, 0, cols/2)
1060                         for hgvsID := range hgvsCols {
1061                                 hgvsIDs = append(hgvsIDs, hgvsID)
1062                         }
1063                         sort.Strings(hgvsIDs)
1064                         var hgvsLabels bytes.Buffer
1065                         for idx, hgvsID := range hgvsIDs {
1066                                 fmt.Fprintf(&hgvsLabels, "%d,%s\n", idx, hgvsID)
1067                                 for ph := 0; ph < 2; ph++ {
1068                                         hgvscol := hgvsCols[hgvsID][ph]
1069                                         for row, val := range hgvscol {
1070                                                 out[row*cols+idx*2+ph] = val
1071                                         }
1072                                 }
1073                         }
1074                         err = writeNumpyInt16(fmt.Sprintf("%s/hgvs.npy", *outputDir), out, rows, cols)
1075                         if err != nil {
1076                                 return err
1077                         }
1078
1079                         fnm := fmt.Sprintf("%s/hgvs.annotations.csv", *outputDir)
1080                         log.Printf("writing hgvs labels: %s", fnm)
1081                         err = ioutil.WriteFile(fnm, hgvsLabels.Bytes(), 0777)
1082                         if err != nil {
1083                                 return err
1084                         }
1085                 }
1086         }
1087         if *onehotSingle || *onlyPCA {
1088                 nzCount := 0
1089                 for _, part := range onehotIndirect {
1090                         nzCount += len(part[0])
1091                 }
1092                 onehot := make([]uint32, nzCount*2) // [r,r,r,...,c,c,c,...]
1093                 var xrefs []onehotXref
1094                 chunkOffset := uint32(0)
1095                 outcol := 0
1096                 for i, part := range onehotIndirect {
1097                         for i := range part[1] {
1098                                 part[1][i] += chunkOffset
1099                         }
1100                         copy(onehot[outcol:], part[0])
1101                         copy(onehot[outcol+nzCount:], part[1])
1102                         xrefs = append(xrefs, onehotXrefs[i]...)
1103
1104                         outcol += len(part[0])
1105                         chunkOffset += onehotChunkSize[i]
1106
1107                         part[0] = nil
1108                         part[1] = nil
1109                         onehotXrefs[i] = nil
1110                         debug.FreeOSMemory()
1111                 }
1112                 if *onehotSingle {
1113                         fnm := fmt.Sprintf("%s/onehot.npy", *outputDir)
1114                         err = writeNumpyUint32(fnm, onehot, 2, nzCount)
1115                         if err != nil {
1116                                 return err
1117                         }
1118                         fnm = fmt.Sprintf("%s/onehot-columns.npy", *outputDir)
1119                         err = writeNumpyInt32(fnm, onehotXref2int32(xrefs), 5, len(xrefs))
1120                         if err != nil {
1121                                 return err
1122                         }
1123                 }
1124                 if *onlyPCA {
1125                         cols := 0
1126                         for _, c := range onehot[nzCount:] {
1127                                 if int(c) >= cols {
1128                                         cols = int(c) + 1
1129                                 }
1130                         }
1131                         if cols == 0 {
1132                                 return fmt.Errorf("cannot do PCA: one-hot matrix is empty")
1133                         }
1134                         log.Printf("have %d one-hot cols", cols)
1135                         stride := 1
1136                         for *maxPCATiles > 0 && cols > *maxPCATiles*2 {
1137                                 cols = (cols + 1) / 2
1138                                 stride = stride * 2
1139                         }
1140                         log.Printf("creating full matrix (%d rows) and training matrix (%d rows) with %d cols, stride %d", len(cmd.cgnames), cmd.trainingSetSize, cols, stride)
1141                         mtxFull := mat.NewDense(len(cmd.cgnames), cols, nil)
1142                         mtxTrain := mat.NewDense(cmd.trainingSetSize, cols, nil)
1143                         for i, c := range onehot[nzCount:] {
1144                                 if int(c/2)%stride == 0 {
1145                                         outcol := int(c/2)/stride*2 + int(c)%2
1146                                         mtxFull.Set(int(onehot[i]), outcol, 1)
1147                                         if trainRow := cmd.trainingSet[int(onehot[i])]; trainRow >= 0 {
1148                                                 mtxTrain.Set(trainRow, outcol, 1)
1149                                         }
1150                                 }
1151                         }
1152                         log.Print("fitting")
1153                         transformer := nlp.NewPCA(*pcaComponents)
1154                         transformer.Fit(mtxTrain.T())
1155                         log.Printf("transforming")
1156                         pca, err := transformer.Transform(mtxFull.T())
1157                         if err != nil {
1158                                 return err
1159                         }
1160                         pca = pca.T()
1161                         outrows, outcols := pca.Dims()
1162                         log.Printf("copying result to numpy output array: %d rows, %d cols", outrows, outcols)
1163                         out := make([]float64, outrows*outcols)
1164                         for i := 0; i < outrows; i++ {
1165                                 for j := 0; j < outcols; j++ {
1166                                         out[i*outcols+j] = pca.At(i, j)
1167                                 }
1168                         }
1169                         fnm := fmt.Sprintf("%s/pca.npy", *outputDir)
1170                         log.Printf("writing numpy: %s", fnm)
1171                         output, err := os.OpenFile(fnm, os.O_CREATE|os.O_TRUNC|os.O_WRONLY, 0777)
1172                         if err != nil {
1173                                 return err
1174                         }
1175                         npw, err := gonpy.NewWriter(nopCloser{output})
1176                         if err != nil {
1177                                 return fmt.Errorf("gonpy.NewWriter: %w", err)
1178                         }
1179                         npw.Shape = []int{outrows, outcols}
1180                         err = npw.WriteFloat64(out)
1181                         if err != nil {
1182                                 return fmt.Errorf("WriteFloat64: %w", err)
1183                         }
1184                         err = output.Close()
1185                         if err != nil {
1186                                 return err
1187                         }
1188                         log.Print("done")
1189                 }
1190         }
1191         if !*mergeOutput && !*onehotChunked && !*onehotSingle && !*onlyPCA {
1192                 tagoffsetFilename := *outputDir + "/chunk-tag-offset.csv"
1193                 log.Infof("writing tag offsets to %s", tagoffsetFilename)
1194                 var f *os.File
1195                 f, err = os.Create(tagoffsetFilename)
1196                 if err != nil {
1197                         return err
1198                 }
1199                 defer f.Close()
1200                 for idx, offset := range chunkStartTag {
1201                         _, err = fmt.Fprintf(f, "%q,%d\n", fmt.Sprintf("matrix.%04d.npy", idx), offset)
1202                         if err != nil {
1203                                 err = fmt.Errorf("write %s: %w", tagoffsetFilename, err)
1204                                 return err
1205                         }
1206                 }
1207                 err = f.Close()
1208                 if err != nil {
1209                         err = fmt.Errorf("close %s: %w", tagoffsetFilename, err)
1210                         return err
1211                 }
1212         }
1213         return nil
1214 }
1215
1216 // Read training set file(s) from path (may be dir or file) and set up
1217 // cmd.trainingSet.
1218 //
1219 // cmd.trainingSet[i] == n >= 0 if cmd.cgnames[i] is the nth training
1220 // set sample.
1221 //
1222 // cmd.trainingSet[i] == -1 if cmd.cgnames[i] is not in the training
1223 // set.
1224 func (cmd *sliceNumpy) loadTrainingSet(path string) error {
1225         cmd.trainingSet = make([]int, len(cmd.cgnames))
1226         if path == "" {
1227                 cmd.trainingSetSize = len(cmd.cgnames)
1228                 for i := range cmd.trainingSet {
1229                         cmd.trainingSet[i] = i
1230                 }
1231                 return nil
1232         }
1233         for i := range cmd.trainingSet {
1234                 cmd.trainingSet[i] = -1
1235         }
1236         infiles, err := allFiles(path, nil)
1237         if err != nil {
1238                 return err
1239         }
1240         for _, infile := range infiles {
1241                 f, err := open(infile)
1242                 if err != nil {
1243                         return err
1244                 }
1245                 buf, err := io.ReadAll(f)
1246                 f.Close()
1247                 if err != nil {
1248                         return err
1249                 }
1250                 for _, tsv := range bytes.Split(buf, []byte{'\n'}) {
1251                         if len(tsv) == 0 {
1252                                 continue
1253                         }
1254                         split := strings.Split(string(tsv), "\t")
1255                         pattern := split[0]
1256                         found := -1
1257                         for i, name := range cmd.cgnames {
1258                                 if strings.Contains(name, pattern) {
1259                                         if found >= 0 {
1260                                                 log.Warnf("pattern %q in %s already matched sample ID %q -- not using %q", pattern, infile, cmd.cgnames[found], name)
1261                                         } else {
1262                                                 found = i
1263                                                 cmd.trainingSet[found] = 1
1264                                         }
1265                                 }
1266                         }
1267                         if found < 0 {
1268                                 log.Warnf("pattern %q in %s does not match any genome IDs", pattern, infile)
1269                                 continue
1270                         }
1271                 }
1272         }
1273         tsi := 0
1274         for i, x := range cmd.trainingSet {
1275                 if x == 1 {
1276                         cmd.trainingSet[i] = tsi
1277                         tsi++
1278                 }
1279         }
1280         cmd.trainingSetSize = tsi + 1
1281         return nil
1282 }
1283
1284 // Read case/control files, remove non-case/control entries from
1285 // cmd.cgnames, and build cmd.chi2Cases.
1286 func (cmd *sliceNumpy) useCaseControlFiles() error {
1287         if cmd.chi2CaseControlFile == "" {
1288                 return nil
1289         }
1290         infiles, err := allFiles(cmd.chi2CaseControlFile, nil)
1291         if err != nil {
1292                 return err
1293         }
1294         // index in cmd.cgnames => case(true) / control(false)
1295         cc := map[int]bool{}
1296         for _, infile := range infiles {
1297                 f, err := open(infile)
1298                 if err != nil {
1299                         return err
1300                 }
1301                 buf, err := io.ReadAll(f)
1302                 f.Close()
1303                 if err != nil {
1304                         return err
1305                 }
1306                 ccCol := -1
1307                 for _, tsv := range bytes.Split(buf, []byte{'\n'}) {
1308                         if len(tsv) == 0 {
1309                                 continue
1310                         }
1311                         split := strings.Split(string(tsv), "\t")
1312                         if ccCol < 0 {
1313                                 // header row
1314                                 for col, name := range split {
1315                                         if name == cmd.chi2CaseControlColumn {
1316                                                 ccCol = col
1317                                                 break
1318                                         }
1319                                 }
1320                                 if ccCol < 0 {
1321                                         return fmt.Errorf("%s: no column named %q in header row %q", infile, cmd.chi2CaseControlColumn, tsv)
1322                                 }
1323                                 continue
1324                         }
1325                         if len(split) <= ccCol {
1326                                 continue
1327                         }
1328                         pattern := split[0]
1329                         found := -1
1330                         for i, name := range cmd.cgnames {
1331                                 if strings.Contains(name, pattern) {
1332                                         if found >= 0 {
1333                                                 log.Warnf("pattern %q in %s matches multiple genome IDs (%q, %q)", pattern, infile, cmd.cgnames[found], name)
1334                                         }
1335                                         found = i
1336                                         if split[ccCol] == "0" {
1337                                                 cc[found] = false
1338                                         }
1339                                         if split[ccCol] == "1" {
1340                                                 cc[found] = true
1341                                         }
1342                                 }
1343                         }
1344                         if found < 0 {
1345                                 log.Warnf("pattern %q in %s does not match any genome IDs", pattern, infile)
1346                                 continue
1347                         }
1348                 }
1349         }
1350         allnames := cmd.cgnames
1351         cmd.cgnames = nil
1352         cmd.chi2Cases = nil
1353         ncases := 0
1354         for i, name := range allnames {
1355                 if cc, ok := cc[i]; ok {
1356                         cmd.cgnames = append(cmd.cgnames, name)
1357                         cmd.chi2Cases = append(cmd.chi2Cases, cc)
1358                         if cc {
1359                                 ncases++
1360                         }
1361                 }
1362         }
1363         log.Printf("%d cases, %d controls, %d neither (dropped)", ncases, len(cmd.cgnames)-ncases, len(allnames)-len(cmd.cgnames))
1364         return nil
1365 }
1366
1367 func (cmd *sliceNumpy) filterHGVScolpair(colpair [2][]int8) bool {
1368         if cmd.chi2PValue >= 1 {
1369                 return true
1370         }
1371         col0 := make([]bool, 0, len(cmd.chi2Cases))
1372         col1 := make([]bool, 0, len(cmd.chi2Cases))
1373         cases := make([]bool, 0, len(cmd.chi2Cases))
1374         for i, c := range cmd.chi2Cases {
1375                 if colpair[0][i] < 0 {
1376                         continue
1377                 }
1378                 col0 = append(col0, colpair[0][i] != 0)
1379                 col1 = append(col1, colpair[1][i] != 0)
1380                 cases = append(cases, c)
1381         }
1382         return len(cases) >= cmd.minCoverage &&
1383                 (pvalue(col0, cases) <= cmd.chi2PValue || pvalue(col1, cases) <= cmd.chi2PValue)
1384 }
1385
1386 func writeNumpyUint32(fnm string, out []uint32, rows, cols int) error {
1387         output, err := os.Create(fnm)
1388         if err != nil {
1389                 return err
1390         }
1391         defer output.Close()
1392         bufw := bufio.NewWriterSize(output, 1<<26)
1393         npw, err := gonpy.NewWriter(nopCloser{bufw})
1394         if err != nil {
1395                 return err
1396         }
1397         log.WithFields(log.Fields{
1398                 "filename": fnm,
1399                 "rows":     rows,
1400                 "cols":     cols,
1401                 "bytes":    rows * cols * 4,
1402         }).Infof("writing numpy: %s", fnm)
1403         npw.Shape = []int{rows, cols}
1404         npw.WriteUint32(out)
1405         err = bufw.Flush()
1406         if err != nil {
1407                 return err
1408         }
1409         return output.Close()
1410 }
1411
1412 func writeNumpyInt32(fnm string, out []int32, rows, cols int) error {
1413         output, err := os.Create(fnm)
1414         if err != nil {
1415                 return err
1416         }
1417         defer output.Close()
1418         bufw := bufio.NewWriterSize(output, 1<<26)
1419         npw, err := gonpy.NewWriter(nopCloser{bufw})
1420         if err != nil {
1421                 return err
1422         }
1423         log.WithFields(log.Fields{
1424                 "filename": fnm,
1425                 "rows":     rows,
1426                 "cols":     cols,
1427                 "bytes":    rows * cols * 4,
1428         }).Infof("writing numpy: %s", fnm)
1429         npw.Shape = []int{rows, cols}
1430         npw.WriteInt32(out)
1431         err = bufw.Flush()
1432         if err != nil {
1433                 return err
1434         }
1435         return output.Close()
1436 }
1437
1438 func writeNumpyInt16(fnm string, out []int16, rows, cols int) error {
1439         output, err := os.Create(fnm)
1440         if err != nil {
1441                 return err
1442         }
1443         defer output.Close()
1444         bufw := bufio.NewWriterSize(output, 1<<26)
1445         npw, err := gonpy.NewWriter(nopCloser{bufw})
1446         if err != nil {
1447                 return err
1448         }
1449         log.WithFields(log.Fields{
1450                 "filename": fnm,
1451                 "rows":     rows,
1452                 "cols":     cols,
1453                 "bytes":    rows * cols * 2,
1454         }).Infof("writing numpy: %s", fnm)
1455         npw.Shape = []int{rows, cols}
1456         npw.WriteInt16(out)
1457         err = bufw.Flush()
1458         if err != nil {
1459                 return err
1460         }
1461         return output.Close()
1462 }
1463
1464 func writeNumpyInt8(fnm string, out []int8, rows, cols int) error {
1465         output, err := os.Create(fnm)
1466         if err != nil {
1467                 return err
1468         }
1469         defer output.Close()
1470         bufw := bufio.NewWriterSize(output, 1<<26)
1471         npw, err := gonpy.NewWriter(nopCloser{bufw})
1472         if err != nil {
1473                 return err
1474         }
1475         log.WithFields(log.Fields{
1476                 "filename": fnm,
1477                 "rows":     rows,
1478                 "cols":     cols,
1479                 "bytes":    rows * cols,
1480         }).Infof("writing numpy: %s", fnm)
1481         npw.Shape = []int{rows, cols}
1482         npw.WriteInt8(out)
1483         err = bufw.Flush()
1484         if err != nil {
1485                 return err
1486         }
1487         return output.Close()
1488 }
1489
1490 func allele2homhet(colpair [2][]int8) {
1491         a, b := colpair[0], colpair[1]
1492         for i, av := range a {
1493                 bv := b[i]
1494                 if av < 0 || bv < 0 {
1495                         // no-call
1496                         a[i], b[i] = -1, -1
1497                 } else if av > 0 && bv > 0 {
1498                         // hom
1499                         a[i], b[i] = 1, 0
1500                 } else if av > 0 || bv > 0 {
1501                         // het
1502                         a[i], b[i] = 0, 1
1503                 } else {
1504                         // ref (or a different variant in same position)
1505                         // (this is a no-op) a[i], b[i] = 0, 0
1506                 }
1507         }
1508 }
1509
1510 type onehotXref struct {
1511         tag     tagID
1512         variant tileVariantID
1513         hom     bool
1514         pvalue  float64
1515 }
1516
1517 const onehotXrefSize = unsafe.Sizeof(onehotXref{})
1518
1519 // Build onehot matrix (m[tileVariantIndex][genome] == 0 or 1) for all
1520 // variants of a single tile/tag#.
1521 //
1522 // Return nil if no tile variant passes Χ² filter.
1523 func (cmd *sliceNumpy) tv2homhet(cgs map[string]CompactGenome, maxv tileVariantID, remap []tileVariantID, tag, chunkstarttag tagID, seq map[tagID][]TileVariant) ([][]int8, []onehotXref) {
1524         if tag == cmd.debugTag {
1525                 tv := make([]tileVariantID, len(cmd.cgnames)*2)
1526                 for i, name := range cmd.cgnames {
1527                         copy(tv[i*2:(i+1)*2], cgs[name].Variants[(tag-chunkstarttag)*2:])
1528                 }
1529                 log.WithFields(logrus.Fields{
1530                         "cgs[i].Variants[tag*2+j]": tv,
1531                         "maxv":                     maxv,
1532                         "remap":                    remap,
1533                         "tag":                      tag,
1534                         "chunkstarttag":            chunkstarttag,
1535                 }).Info("tv2homhet()")
1536         }
1537         if maxv < 1 || (maxv < 2 && !cmd.includeVariant1) {
1538                 // everyone has the most common variant (of the variants we don't drop)
1539                 return nil, nil
1540         }
1541         tagoffset := tag - chunkstarttag
1542         coverage := 0
1543         for _, cg := range cgs {
1544                 alleles := 0
1545                 for _, v := range cg.Variants[tagoffset*2 : tagoffset*2+2] {
1546                         if v > 0 && int(v) < len(seq[tag]) && len(seq[tag][v].Sequence) > 0 {
1547                                 alleles++
1548                         }
1549                 }
1550                 if alleles == 2 {
1551                         coverage++
1552                 }
1553         }
1554         if coverage < cmd.minCoverage {
1555                 return nil, nil
1556         }
1557         obs := make([][]bool, (maxv+1)*2) // 2 slices (hom + het) for each variant#
1558         for i := range obs {
1559                 obs[i] = make([]bool, len(cmd.cgnames))
1560         }
1561         for cgid, name := range cmd.cgnames {
1562                 cgvars := cgs[name].Variants[tagoffset*2:]
1563                 tv0, tv1 := remap[cgvars[0]], remap[cgvars[1]]
1564                 for v := tileVariantID(1); v <= maxv; v++ {
1565                         if tv0 == v && tv1 == v {
1566                                 obs[v*2][cgid] = true
1567                         } else if tv0 == v || tv1 == v {
1568                                 obs[v*2+1][cgid] = true
1569                         }
1570                 }
1571         }
1572         var onehot [][]int8
1573         var xref []onehotXref
1574         for col := 2; col < len(obs); col++ {
1575                 // col 0,1 correspond to tile variant 0, i.e.,
1576                 // no-call; col 2,3 correspond to the most common
1577                 // variant; so we (normally) start at col 4.
1578                 if col < 4 && !cmd.includeVariant1 {
1579                         continue
1580                 }
1581                 p := pvalue(obs[col], cmd.chi2Cases)
1582                 if cmd.chi2PValue < 1 && !(p < cmd.chi2PValue) {
1583                         continue
1584                 }
1585                 onehot = append(onehot, bool2int8(obs[col]))
1586                 xref = append(xref, onehotXref{
1587                         tag:     tag,
1588                         variant: tileVariantID(col >> 1),
1589                         hom:     col&1 == 0,
1590                         pvalue:  p,
1591                 })
1592         }
1593         return onehot, xref
1594 }
1595
1596 func bool2int8(in []bool) []int8 {
1597         out := make([]int8, len(in))
1598         for i, v := range in {
1599                 if v {
1600                         out[i] = 1
1601                 }
1602         }
1603         return out
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 }