slicenumpy.go

   1 // Copyright (C) The Lightning Authors. All rights reserved.
   2 //
   3 // SPDX-License-Identifier: AGPL-3.0
   4
   5 package lightning
   6
   7 import (
   8         "bufio"
   9         "bytes"
  10         "encoding/gob"
  11         "errors"
  12         "flag"
  13         "fmt"
  14         "io"
  15         "io/ioutil"
  16         "math"
  17         "net/http"
  18         _ "net/http/pprof"
  19         "os"
  20         "regexp"
  21         "runtime"
  22         "runtime/debug"
  23         "sort"
  24         "strconv"
  25         "strings"
  26         "sync/atomic"
  27         "unsafe"
  28
  29         "git.arvados.org/arvados.git/sdk/go/arvados"
  30         "github.com/arvados/lightning/hgvs"
  31         "github.com/james-bowman/nlp"
  32         "github.com/kshedden/gonpy"
  33         "github.com/sirupsen/logrus"
  34         log "github.com/sirupsen/logrus"
  35         "golang.org/x/crypto/blake2b"
  36         "gonum.org/v1/gonum/mat"
  37 )
  38
  39 const annotationMaxTileSpan = 100
  40
  41 type sliceNumpy struct {
  42         filter          filter
  43         threads         int
  44         chi2Cases       []bool
  45         chi2PValue      float64
  46         minCoverage     int
  47         includeVariant1 bool
  48         debugTag        tagID
  49
  50         cgnames         []string
  51         samples         []sampleInfo
  52         trainingSet     []int // samples index => training set index, or -1 if not in training set
  53         trainingSetSize int
  54 }
  55
  56 func (cmd *sliceNumpy) RunCommand(prog string, args []string, stdin io.Reader, stdout, stderr io.Writer) int {
  57         err := cmd.run(prog, args, stdin, stdout, stderr)
  58         if err != nil {
  59                 fmt.Fprintf(stderr, "%s\n", err)
  60                 return 1
  61         }
  62         return 0
  63 }
  64
  65 func (cmd *sliceNumpy) run(prog string, args []string, stdin io.Reader, stdout, stderr io.Writer) error {
  66         flags := flag.NewFlagSet("", flag.ContinueOnError)
  67         flags.SetOutput(stderr)
  68         pprof := flags.String("pprof", "", "serve Go profile data at http://`[addr]:port`")
  69         runlocal := flags.Bool("local", false, "run on local host (default: run in an arvados container)")
  70         arvadosRAM := flags.Int("arvados-ram", 750000000000, "amount of memory to request for arvados container (`bytes`)")
  71         arvadosVCPUs := flags.Int("arvados-vcpus", 96, "number of VCPUs to request for arvados container")
  72         projectUUID := flags.String("project", "", "project `UUID` for output data")
  73         priority := flags.Int("priority", 500, "container request priority")
  74         inputDir := flags.String("input-dir", "./in", "input `directory`")
  75         outputDir := flags.String("output-dir", "./out", "output `directory`")
  76         ref := flags.String("ref", "", "reference name (if blank, choose last one that appears in input)")
  77         regionsFilename := flags.String("regions", "", "only output columns/annotations that intersect regions in specified bed `file`")
  78         expandRegions := flags.Int("expand-regions", 0, "expand specified regions by `N` base pairs on each side`")
  79         mergeOutput := flags.Bool("merge-output", false, "merge output into one matrix.npy and one matrix.annotations.csv")
  80         hgvsSingle := flags.Bool("single-hgvs-matrix", false, "also generate hgvs-based matrix")
  81         hgvsChunked := flags.Bool("chunked-hgvs-matrix", false, "also generate hgvs-based matrix per chromosome")
  82         onehotSingle := flags.Bool("single-onehot", false, "generate one-hot tile-based matrix")
  83         onehotChunked := flags.Bool("chunked-onehot", false, "generate one-hot tile-based matrix per input chunk")
  84         samplesFilename := flags.String("samples", "", "`samples.csv` file with training/validation and case/control groups (see 'lightning choose-samples')")
  85         caseControlOnly := flags.Bool("case-control-only", false, "drop samples that are not in case/control groups")
  86         onlyPCA := flags.Bool("pca", false, "generate pca matrix")
  87         pcaComponents := flags.Int("pca-components", 4, "number of PCA components")
  88         maxPCATiles := flags.Int("max-pca-tiles", 0, "maximum tiles to use as PCA input (filter, then drop every 2nd colum pair until below max)")
  89         debugTag := flags.Int("debug-tag", -1, "log debugging details about specified tag")
  90         flags.IntVar(&cmd.threads, "threads", 16, "number of memory-hungry assembly threads, and number of VCPUs to request for arvados container")
  91         flags.Float64Var(&cmd.chi2PValue, "chi2-p-value", 1, "do Χ² test and omit columns with p-value above this threshold")
  92         flags.BoolVar(&cmd.includeVariant1, "include-variant-1", false, "include most common variant when building one-hot matrix")
  93         cmd.filter.Flags(flags)
  94         err := flags.Parse(args)
  95         if err == flag.ErrHelp {
  96                 return nil
  97         } else if err != nil {
  98                 return err
  99         }
 100
 101         if *pprof != "" {
 102                 go func() {
 103                         log.Println(http.ListenAndServe(*pprof, nil))
 104                 }()
 105         }
 106
 107         if cmd.chi2PValue != 1 && *samplesFilename == "" {
 108                 return fmt.Errorf("cannot use provided -chi2-p-value=%f because -samples= value is empty", cmd.chi2PValue)
 109         }
 110
 111         cmd.debugTag = tagID(*debugTag)
 112
 113         if !*runlocal {
 114                 runner := arvadosContainerRunner{
 115                         Name:        "lightning slice-numpy",
 116                         Client:      arvados.NewClientFromEnv(),
 117                         ProjectUUID: *projectUUID,
 118                         RAM:         int64(*arvadosRAM),
 119                         VCPUs:       *arvadosVCPUs,
 120                         Priority:    *priority,
 121                         KeepCache:   2,
 122                         APIAccess:   true,
 123                 }
 124                 err = runner.TranslatePaths(inputDir, regionsFilename, samplesFilename)
 125                 if err != nil {
 126                         return err
 127                 }
 128                 runner.Args = []string{"slice-numpy", "-local=true",
 129                         "-pprof=:6060",
 130                         "-input-dir=" + *inputDir,
 131                         "-output-dir=/mnt/output",
 132                         "-threads=" + fmt.Sprintf("%d", cmd.threads),
 133                         "-regions=" + *regionsFilename,
 134                         "-expand-regions=" + fmt.Sprintf("%d", *expandRegions),
 135                         "-merge-output=" + fmt.Sprintf("%v", *mergeOutput),
 136                         "-single-hgvs-matrix=" + fmt.Sprintf("%v", *hgvsSingle),
 137                         "-chunked-hgvs-matrix=" + fmt.Sprintf("%v", *hgvsChunked),
 138                         "-single-onehot=" + fmt.Sprintf("%v", *onehotSingle),
 139                         "-chunked-onehot=" + fmt.Sprintf("%v", *onehotChunked),
 140                         "-samples=" + *samplesFilename,
 141                         "-case-control-only=" + fmt.Sprintf("%v", *caseControlOnly),
 142                         "-pca=" + fmt.Sprintf("%v", *onlyPCA),
 143                         "-pca-components=" + fmt.Sprintf("%d", *pcaComponents),
 144                         "-max-pca-tiles=" + fmt.Sprintf("%d", *maxPCATiles),
 145                         "-chi2-p-value=" + fmt.Sprintf("%f", cmd.chi2PValue),
 146                         "-include-variant-1=" + fmt.Sprintf("%v", cmd.includeVariant1),
 147                         "-debug-tag=" + fmt.Sprintf("%d", cmd.debugTag),
 148                 }
 149                 runner.Args = append(runner.Args, cmd.filter.Args()...)
 150                 var output string
 151                 output, err = runner.Run()
 152                 if err != nil {
 153                         return err
 154                 }
 155                 fmt.Fprintln(stdout, output)
 156                 return nil
 157         }
 158
 159         infiles, err := allFiles(*inputDir, matchGobFile)
 160         if err != nil {
 161                 return err
 162         }
 163         if len(infiles) == 0 {
 164                 err = fmt.Errorf("no input files found in %s", *inputDir)
 165                 return err
 166         }
 167         sort.Strings(infiles)
 168
 169         var refseq map[string][]tileLibRef
 170         var reftiledata = make(map[tileLibRef][]byte, 11000000)
 171         in0, err := open(infiles[0])
 172         if err != nil {
 173                 return err
 174         }
 175
 176         matchGenome, err := regexp.Compile(cmd.filter.MatchGenome)
 177         if err != nil {
 178                 err = fmt.Errorf("-match-genome: invalid regexp: %q", cmd.filter.MatchGenome)
 179                 return err
 180         }
 181
 182         if *samplesFilename != "" {
 183                 cmd.samples, err = cmd.loadSampleInfo(*samplesFilename)
 184                 if err != nil {
 185                         return err
 186                 }
 187         } else if *caseControlOnly {
 188                 return fmt.Errorf("-case-control-only does not make sense without -samples")
 189         }
 190
 191         cmd.cgnames = nil
 192         var tagset [][]byte
 193         err = DecodeLibrary(in0, strings.HasSuffix(infiles[0], ".gz"), func(ent *LibraryEntry) error {
 194                 if len(ent.TagSet) > 0 {
 195                         tagset = ent.TagSet
 196                 }
 197                 for _, cseq := range ent.CompactSequences {
 198                         if cseq.Name == *ref || *ref == "" {
 199                                 refseq = cseq.TileSequences
 200                         }
 201                 }
 202                 for _, cg := range ent.CompactGenomes {
 203                         if matchGenome.MatchString(cg.Name) {
 204                                 cmd.cgnames = append(cmd.cgnames, cg.Name)
 205                         }
 206                 }
 207                 for _, tv := range ent.TileVariants {
 208                         if tv.Ref {
 209                                 reftiledata[tileLibRef{tv.Tag, tv.Variant}] = tv.Sequence
 210                         }
 211                 }
 212                 return nil
 213         })
 214         if err != nil {
 215                 return err
 216         }
 217         in0.Close()
 218         if refseq == nil {
 219                 err = fmt.Errorf("%s: reference sequence not found", infiles[0])
 220                 return err
 221         }
 222         if len(tagset) == 0 {
 223                 err = fmt.Errorf("tagset not found")
 224                 return err
 225         }
 226
 227         taglib := &tagLibrary{}
 228         err = taglib.setTags(tagset)
 229         if err != nil {
 230                 return err
 231         }
 232         taglen := taglib.TagLen()
 233         sort.Strings(cmd.cgnames)
 234
 235         if len(cmd.cgnames) == 0 {
 236                 return fmt.Errorf("fatal: 0 matching samples in library, nothing to do")
 237         }
 238         cmd.trainingSet = make([]int, len(cmd.cgnames))
 239         if *samplesFilename == "" {
 240                 cmd.trainingSetSize = len(cmd.cgnames)
 241                 for i, name := range cmd.cgnames {
 242                         cmd.samples = append(cmd.samples, sampleInfo{
 243                                 id:         trimFilenameForLabel(name),
 244                                 isTraining: true,
 245                         })
 246                         cmd.trainingSet[i] = i
 247                 }
 248         } else if len(cmd.cgnames) != len(cmd.samples) {
 249                 return fmt.Errorf("mismatched sample list: %d samples in library, %d in %s", len(cmd.cgnames), len(cmd.samples), *samplesFilename)
 250         } else {
 251                 for i, name := range cmd.cgnames {
 252                         if s := trimFilenameForLabel(name); s != cmd.samples[i].id {
 253                                 return fmt.Errorf("mismatched sample list: sample %d is %q in library, %q in %s", i, s, cmd.samples[i].id, *samplesFilename)
 254                         }
 255                 }
 256                 if *caseControlOnly {
 257                         for i := 0; i < len(cmd.samples); i++ {
 258                                 if !cmd.samples[i].isTraining && !cmd.samples[i].isValidation {
 259                                         if i+1 < len(cmd.samples) {
 260                                                 copy(cmd.samples[i:], cmd.samples[i+1:])
 261                                                 copy(cmd.cgnames[i:], cmd.cgnames[i+1:])
 262                                         }
 263                                         cmd.samples = cmd.samples[:len(cmd.samples)-1]
 264                                         cmd.cgnames = cmd.cgnames[:len(cmd.cgnames)-1]
 265                                         i--
 266                                 }
 267                         }
 268                 }
 269                 cmd.chi2Cases = nil
 270                 cmd.trainingSetSize = 0
 271                 for i := range cmd.cgnames {
 272                         if cmd.samples[i].isTraining {
 273                                 cmd.trainingSet[i] = cmd.trainingSetSize
 274                                 cmd.trainingSetSize++
 275                                 cmd.chi2Cases = append(cmd.chi2Cases, cmd.samples[i].isCase)
 276                         } else {
 277                                 cmd.trainingSet[i] = -1
 278                         }
 279                 }
 280         }
 281         if cmd.filter.MinCoverage == 1 {
 282                 // In the generic formula below, floating point
 283                 // arithmetic can effectively push the coverage
 284                 // threshold above 1.0, which is impossible/useless.
 285                 // 1.0 needs to mean exactly 100% coverage.
 286                 cmd.minCoverage = len(cmd.cgnames)
 287         } else {
 288                 cmd.minCoverage = int(math.Ceil(cmd.filter.MinCoverage * float64(len(cmd.cgnames))))
 289         }
 290
 291         {
 292                 samplesOutFilename := *outputDir + "/samples.csv"
 293                 log.Infof("writing sample metadata to %s", samplesOutFilename)
 294                 var f *os.File
 295                 f, err = os.Create(samplesOutFilename)
 296                 if err != nil {
 297                         return err
 298                 }
 299                 defer f.Close()
 300                 for i, si := range cmd.samples {
 301                         var cc, tv string
 302                         if si.isCase {
 303                                 cc = "1"
 304                         } else if si.isControl {
 305                                 cc = "0"
 306                         }
 307                         if si.isTraining {
 308                                 tv = "1"
 309                         } else {
 310                                 tv = "0"
 311                         }
 312                         _, err = fmt.Fprintf(f, "%d,%s,%s,%s\n", i, si.id, cc, tv)
 313                         if err != nil {
 314                                 err = fmt.Errorf("write %s: %w", samplesOutFilename, err)
 315                                 return err
 316                         }
 317                 }
 318                 err = f.Close()
 319                 if err != nil {
 320                         err = fmt.Errorf("close %s: %w", samplesOutFilename, err)
 321                         return err
 322                 }
 323                 log.Print("done")
 324         }
 325
 326         log.Info("indexing reference tiles")
 327         type reftileinfo struct {
 328                 variant  tileVariantID
 329                 seqname  string // chr1
 330                 pos      int    // distance from start of chromosome to starttag
 331                 tiledata []byte // acgtggcaa...
 332                 excluded bool   // true if excluded by regions file
 333                 nexttag  tagID  // tagID of following tile (-1 for last tag of chromosome)
 334         }
 335         isdup := map[tagID]bool{}
 336         reftile := map[tagID]*reftileinfo{}
 337         for seqname, cseq := range refseq {
 338                 pos := 0
 339                 lastreftag := tagID(-1)
 340                 for _, libref := range cseq {
 341                         if cmd.filter.MaxTag >= 0 && libref.Tag > tagID(cmd.filter.MaxTag) {
 342                                 continue
 343                         }
 344                         tiledata := reftiledata[libref]
 345                         if len(tiledata) == 0 {
 346                                 err = fmt.Errorf("missing tiledata for tag %d variant %d in %s in ref", libref.Tag, libref.Variant, seqname)
 347                                 return err
 348                         }
 349                         foundthistag := false
 350                         taglib.FindAll(tiledata[:len(tiledata)-1], func(tagid tagID, offset, _ int) {
 351                                 if !foundthistag && tagid == libref.Tag {
 352                                         foundthistag = true
 353                                         return
 354                                 }
 355                                 if dupref, ok := reftile[tagid]; ok {
 356                                         log.Printf("dropping reference tile %+v from %s @ %d, tag not unique, also found inside %+v from %s @ %d", tileLibRef{Tag: tagid, Variant: dupref.variant}, dupref.seqname, dupref.pos, libref, seqname, pos+offset+1)
 357                                         delete(reftile, tagid)
 358                                 } else {
 359                                         log.Printf("found tag %d at offset %d inside tile variant %+v on %s @ %d", tagid, offset, libref, seqname, pos+offset+1)
 360                                 }
 361                                 isdup[tagid] = true
 362                         })
 363                         if isdup[libref.Tag] {
 364                                 log.Printf("dropping reference tile %+v from %s @ %d, tag not unique", libref, seqname, pos)
 365                         } else if reftile[libref.Tag] != nil {
 366                                 log.Printf("dropping reference tile %+v from %s @ %d, tag not unique", tileLibRef{Tag: libref.Tag, Variant: reftile[libref.Tag].variant}, reftile[libref.Tag].seqname, reftile[libref.Tag].pos)
 367                                 delete(reftile, libref.Tag)
 368                                 log.Printf("dropping reference tile %+v from %s @ %d, tag not unique", libref, seqname, pos)
 369                                 isdup[libref.Tag] = true
 370                         } else {
 371                                 reftile[libref.Tag] = &reftileinfo{
 372                                         seqname:  seqname,
 373                                         variant:  libref.Variant,
 374                                         tiledata: tiledata,
 375                                         pos:      pos,
 376                                         nexttag:  -1,
 377                                 }
 378                                 if lastreftag >= 0 {
 379                                         reftile[lastreftag].nexttag = libref.Tag
 380                                 }
 381                                 lastreftag = libref.Tag
 382                         }
 383                         pos += len(tiledata) - taglen
 384                 }
 385                 log.Printf("... %s done, len %d", seqname, pos+taglen)
 386         }
 387
 388         var mask *mask
 389         if *regionsFilename != "" {
 390                 log.Printf("loading regions from %s", *regionsFilename)
 391                 mask, err = makeMask(*regionsFilename, *expandRegions)
 392                 if err != nil {
 393                         return err
 394                 }
 395                 log.Printf("before applying mask, len(reftile) == %d", len(reftile))
 396                 log.Printf("deleting reftile entries for regions outside %d intervals", mask.Len())
 397                 for _, rt := range reftile {
 398                         if !mask.Check(strings.TrimPrefix(rt.seqname, "chr"), rt.pos, rt.pos+len(rt.tiledata)) {
 399                                 rt.excluded = true
 400                         }
 401                 }
 402                 log.Printf("after applying mask, len(reftile) == %d", len(reftile))
 403         }
 404
 405         type hgvsColSet map[hgvs.Variant][2][]int8
 406         encodeHGVS := throttle{Max: len(refseq)}
 407         encodeHGVSTodo := map[string]chan hgvsColSet{}
 408         tmpHGVSCols := map[string]*os.File{}
 409         if *hgvsChunked {
 410                 for seqname := range refseq {
 411                         var f *os.File
 412                         f, err = os.Create(*outputDir + "/tmp." + seqname + ".gob")
 413                         if err != nil {
 414                                 return err
 415                         }
 416                         defer os.Remove(f.Name())
 417                         bufw := bufio.NewWriterSize(f, 1<<24)
 418                         enc := gob.NewEncoder(bufw)
 419                         tmpHGVSCols[seqname] = f
 420                         todo := make(chan hgvsColSet, 128)
 421                         encodeHGVSTodo[seqname] = todo
 422                         encodeHGVS.Go(func() error {
 423                                 for colset := range todo {
 424                                         err := enc.Encode(colset)
 425                                         if err != nil {
 426                                                 encodeHGVS.Report(err)
 427                                                 for range todo {
 428                                                 }
 429                                                 return err
 430                                         }
 431                                 }
 432                                 return bufw.Flush()
 433                         })
 434                 }
 435         }
 436
 437         var toMerge [][]int16
 438         if *mergeOutput || *hgvsSingle {
 439                 toMerge = make([][]int16, len(infiles))
 440         }
 441         var onehotIndirect [][2][]uint32 // [chunkIndex][axis][index]
 442         var onehotChunkSize []uint32
 443         var onehotXrefs [][]onehotXref
 444         if *onehotSingle || *onlyPCA {
 445                 onehotIndirect = make([][2][]uint32, len(infiles))
 446                 onehotChunkSize = make([]uint32, len(infiles))
 447                 onehotXrefs = make([][]onehotXref, len(infiles))
 448         }
 449         chunkStartTag := make([]tagID, len(infiles))
 450
 451         throttleMem := throttle{Max: cmd.threads} // TODO: estimate using mem and data size
 452         throttleNumpyMem := throttle{Max: cmd.threads/2 + 1}
 453         log.Info("generating annotations and numpy matrix for each slice")
 454         var errSkip = errors.New("skip infile")
 455         var done int64
 456         for infileIdx, infile := range infiles {
 457                 infileIdx, infile := infileIdx, infile
 458                 throttleMem.Go(func() error {
 459                         seq := make(map[tagID][]TileVariant, 50000)
 460                         cgs := make(map[string]CompactGenome, len(cmd.cgnames))
 461                         f, err := open(infile)
 462                         if err != nil {
 463                                 return err
 464                         }
 465                         defer f.Close()
 466                         log.Infof("%04d: reading %s", infileIdx, infile)
 467                         err = DecodeLibrary(f, strings.HasSuffix(infile, ".gz"), func(ent *LibraryEntry) error {
 468                                 for _, tv := range ent.TileVariants {
 469                                         if tv.Ref {
 470                                                 continue
 471                                         }
 472                                         // Skip tile with no
 473                                         // corresponding ref tile, if
 474                                         // mask is in play (we can't
 475                                         // determine coordinates for
 476                                         // these)
 477                                         if mask != nil && reftile[tv.Tag] == nil {
 478                                                 continue
 479                                         }
 480                                         // Skip tile whose
 481                                         // corresponding ref tile is
 482                                         // outside target regions --
 483                                         // unless it's a potential
 484                                         // spanning tile.
 485                                         if mask != nil && reftile[tv.Tag].excluded &&
 486                                                 (int(tv.Tag+1) >= len(tagset) ||
 487                                                         (bytes.HasSuffix(tv.Sequence, tagset[tv.Tag+1]) && reftile[tv.Tag+1] != nil && !reftile[tv.Tag+1].excluded)) {
 488                                                 continue
 489                                         }
 490                                         if tv.Tag == cmd.debugTag {
 491                                                 log.Printf("infile %d %s tag %d variant %d hash %x", infileIdx, infile, tv.Tag, tv.Variant, tv.Blake2b[:3])
 492                                         }
 493                                         variants := seq[tv.Tag]
 494                                         if len(variants) == 0 {
 495                                                 variants = make([]TileVariant, 100)
 496                                         }
 497                                         for len(variants) <= int(tv.Variant) {
 498                                                 variants = append(variants, TileVariant{})
 499                                         }
 500                                         variants[int(tv.Variant)] = tv
 501                                         seq[tv.Tag] = variants
 502                                 }
 503                                 for _, cg := range ent.CompactGenomes {
 504                                         if cmd.filter.MaxTag >= 0 && cg.StartTag > tagID(cmd.filter.MaxTag) {
 505                                                 return errSkip
 506                                         }
 507                                         if !matchGenome.MatchString(cg.Name) {
 508                                                 continue
 509                                         }
 510                                         // pad to full slice size
 511                                         // to avoid out-of-bounds
 512                                         // checks later
 513                                         if sliceSize := 2 * int(cg.EndTag-cg.StartTag); len(cg.Variants) < sliceSize {
 514                                                 cg.Variants = append(cg.Variants, make([]tileVariantID, sliceSize-len(cg.Variants))...)
 515                                         }
 516                                         cgs[cg.Name] = cg
 517                                 }
 518                                 return nil
 519                         })
 520                         if err == errSkip {
 521                                 return nil
 522                         } else if err != nil {
 523                                 return fmt.Errorf("%04d: DecodeLibrary(%s): err", infileIdx, infile)
 524                         }
 525                         tagstart := cgs[cmd.cgnames[0]].StartTag
 526                         tagend := cgs[cmd.cgnames[0]].EndTag
 527                         chunkStartTag[infileIdx] = tagstart
 528
 529                         // TODO: filters
 530
 531                         log.Infof("%04d: renumber/dedup variants for tags %d-%d", infileIdx, tagstart, tagend)
 532                         variantRemap := make([][]tileVariantID, tagend-tagstart)
 533                         throttleCPU := throttle{Max: runtime.GOMAXPROCS(0)}
 534                         for tag, variants := range seq {
 535                                 tag, variants := tag, variants
 536                                 throttleCPU.Go(func() error {
 537                                         alleleCoverage := 0
 538                                         count := make(map[[blake2b.Size256]byte]int, len(variants))
 539
 540                                         rt := reftile[tag]
 541                                         if rt != nil {
 542                                                 count[blake2b.Sum256(rt.tiledata)] = 0
 543                                         }
 544
 545                                         for cgname, cg := range cgs {
 546                                                 idx := int(tag-tagstart) * 2
 547                                                 for allele := 0; allele < 2; allele++ {
 548                                                         v := cg.Variants[idx+allele]
 549                                                         if v > 0 && len(variants[v].Sequence) > 0 {
 550                                                                 count[variants[v].Blake2b]++
 551                                                                 alleleCoverage++
 552                                                         }
 553                                                         if v > 0 && tag == cmd.debugTag {
 554                                                                 log.Printf("tag %d cg %s allele %d tv %d hash %x count is now %d", tag, cgname, allele, v, variants[v].Blake2b[:3], count[variants[v].Blake2b])
 555                                                         }
 556                                                 }
 557                                         }
 558                                         if alleleCoverage < cmd.minCoverage*2 {
 559                                                 idx := int(tag-tagstart) * 2
 560                                                 for _, cg := range cgs {
 561                                                         cg.Variants[idx] = 0
 562                                                         cg.Variants[idx+1] = 0
 563                                                 }
 564                                                 if tag == cmd.debugTag {
 565                                                         log.Printf("tag %d alleleCoverage %d < min %d, sample data wiped", tag, alleleCoverage, cmd.minCoverage*2)
 566                                                 }
 567                                                 return nil
 568                                         }
 569
 570                                         // hash[i] will be the hash of
 571                                         // the variant(s) that should
 572                                         // be at rank i (0-based).
 573                                         hash := make([][blake2b.Size256]byte, 0, len(count))
 574                                         for b := range count {
 575                                                 hash = append(hash, b)
 576                                         }
 577                                         sort.Slice(hash, func(i, j int) bool {
 578                                                 bi, bj := &hash[i], &hash[j]
 579                                                 if ci, cj := count[*bi], count[*bj]; ci != cj {
 580                                                         return ci > cj
 581                                                 } else {
 582                                                         return bytes.Compare((*bi)[:], (*bj)[:]) < 0
 583                                                 }
 584                                         })
 585                                         // rank[b] will be the 1-based
 586                                         // new variant number for
 587                                         // variants whose hash is b.
 588                                         rank := make(map[[blake2b.Size256]byte]tileVariantID, len(hash))
 589                                         for i, h := range hash {
 590                                                 rank[h] = tileVariantID(i + 1)
 591                                         }
 592                                         if tag == cmd.debugTag {
 593                                                 for h, r := range rank {
 594                                                         log.Printf("tag %d rank(%x) = %v", tag, h[:3], r)
 595                                                 }
 596                                         }
 597                                         // remap[v] will be the new
 598                                         // variant number for original
 599                                         // variant number v.
 600                                         remap := make([]tileVariantID, len(variants))
 601                                         for i, tv := range variants {
 602                                                 remap[i] = rank[tv.Blake2b]
 603                                         }
 604                                         if tag == cmd.debugTag {
 605                                                 for in, out := range remap {
 606                                                         if out > 0 {
 607                                                                 log.Printf("tag %d remap %d => %d", tag, in, out)
 608                                                         }
 609                                                 }
 610                                         }
 611                                         variantRemap[tag-tagstart] = remap
 612                                         if rt != nil {
 613                                                 refrank := rank[blake2b.Sum256(rt.tiledata)]
 614                                                 if tag == cmd.debugTag {
 615                                                         log.Printf("tag %d reftile variant %d => %d", tag, rt.variant, refrank)
 616                                                 }
 617                                                 rt.variant = refrank
 618                                         }
 619                                         return nil
 620                                 })
 621                         }
 622                         throttleCPU.Wait()
 623
 624                         var onehotChunk [][]int8
 625                         var onehotXref []onehotXref
 626
 627                         var annotationsFilename string
 628                         if *onlyPCA {
 629                                 annotationsFilename = "/dev/null"
 630                         } else {
 631                                 annotationsFilename = fmt.Sprintf("%s/matrix.%04d.annotations.csv", *outputDir, infileIdx)
 632                                 log.Infof("%04d: writing %s", infileIdx, annotationsFilename)
 633                         }
 634                         annof, err := os.Create(annotationsFilename)
 635                         if err != nil {
 636                                 return err
 637                         }
 638                         annow := bufio.NewWriterSize(annof, 1<<20)
 639                         outcol := 0
 640                         for tag := tagstart; tag < tagend; tag++ {
 641                                 rt := reftile[tag]
 642                                 if rt == nil && mask != nil {
 643                                         // With no ref tile, we don't
 644                                         // have coordinates to say
 645                                         // this is in the desired
 646                                         // regions -- so it's not.
 647                                         // TODO: handle ref spanning
 648                                         // tile case.
 649                                         continue
 650                                 }
 651                                 if rt != nil && rt.excluded {
 652                                         // TODO: don't skip yet --
 653                                         // first check for spanning
 654                                         // tile variants that
 655                                         // intersect non-excluded ref
 656                                         // tiles.
 657                                         continue
 658                                 }
 659                                 if cmd.filter.MaxTag >= 0 && tag > tagID(cmd.filter.MaxTag) {
 660                                         break
 661                                 }
 662                                 remap := variantRemap[tag-tagstart]
 663                                 if remap == nil {
 664                                         // was not assigned above,
 665                                         // because minCoverage
 666                                         continue
 667                                 }
 668                                 maxv := tileVariantID(0)
 669                                 for _, v := range remap {
 670                                         if maxv < v {
 671                                                 maxv = v
 672                                         }
 673                                 }
 674                                 if *onehotChunked || *onehotSingle || *onlyPCA {
 675                                         onehot, xrefs := cmd.tv2homhet(cgs, maxv, remap, tag, tagstart, seq)
 676                                         if tag == cmd.debugTag {
 677                                                 log.WithFields(logrus.Fields{
 678                                                         "onehot": onehot,
 679                                                         "xrefs":  xrefs,
 680                                                 }).Info("tv2homhet()")
 681                                         }
 682                                         onehotChunk = append(onehotChunk, onehot...)
 683                                         onehotXref = append(onehotXref, xrefs...)
 684                                 }
 685                                 if *onlyPCA {
 686                                         outcol++
 687                                         continue
 688                                 }
 689                                 if rt == nil {
 690                                         // Reference does not use any
 691                                         // variant of this tile
 692                                         //
 693                                         // TODO: diff against the
 694                                         // relevant portion of the
 695                                         // ref's spanning tile
 696                                         outcol++
 697                                         continue
 698                                 }
 699                                 fmt.Fprintf(annow, "%d,%d,%d,=,%s,%d,,,\n", tag, outcol, rt.variant, rt.seqname, rt.pos)
 700                                 variants := seq[tag]
 701                                 reftilestr := strings.ToUpper(string(rt.tiledata))
 702
 703                                 done := make([]bool, maxv+1)
 704                                 variantDiffs := make([][]hgvs.Variant, maxv+1)
 705                                 for v, tv := range variants {
 706                                         v := remap[v]
 707                                         if v == 0 || v == rt.variant || done[v] {
 708                                                 continue
 709                                         } else {
 710                                                 done[v] = true
 711                                         }
 712                                         if len(tv.Sequence) < taglen {
 713                                                 continue
 714                                         }
 715                                         // if reftilestr doesn't end
 716                                         // in the same tag as tv,
 717                                         // extend reftilestr with
 718                                         // following ref tiles until
 719                                         // it does (up to an arbitrary
 720                                         // sanity-check limit)
 721                                         reftilestr := reftilestr
 722                                         endtagstr := strings.ToUpper(string(tv.Sequence[len(tv.Sequence)-taglen:]))
 723                                         for i, rt := 0, rt; i < annotationMaxTileSpan && !strings.HasSuffix(reftilestr, endtagstr) && rt.nexttag >= 0; i++ {
 724                                                 rt = reftile[rt.nexttag]
 725                                                 if rt == nil {
 726                                                         break
 727                                                 }
 728                                                 reftilestr += strings.ToUpper(string(rt.tiledata[taglen:]))
 729                                         }
 730                                         if mask != nil && !mask.Check(strings.TrimPrefix(rt.seqname, "chr"), rt.pos, rt.pos+len(reftilestr)) {
 731                                                 continue
 732                                         }
 733                                         if !strings.HasSuffix(reftilestr, endtagstr) {
 734                                                 fmt.Fprintf(annow, "%d,%d,%d,,%s,%d,,,\n", tag, outcol, v, rt.seqname, rt.pos)
 735                                                 continue
 736                                         }
 737                                         if lendiff := len(reftilestr) - len(tv.Sequence); lendiff < -1000 || lendiff > 1000 {
 738                                                 fmt.Fprintf(annow, "%d,%d,%d,,%s,%d,,,\n", tag, outcol, v, rt.seqname, rt.pos)
 739                                                 continue
 740                                         }
 741                                         diffs, _ := hgvs.Diff(reftilestr, strings.ToUpper(string(tv.Sequence)), 0)
 742                                         for i := range diffs {
 743                                                 diffs[i].Position += rt.pos
 744                                         }
 745                                         for _, diff := range diffs {
 746                                                 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)
 747                                         }
 748                                         if *hgvsChunked {
 749                                                 variantDiffs[v] = diffs
 750                                         }
 751                                 }
 752                                 if *hgvsChunked {
 753                                         // We can now determine, for each HGVS
 754                                         // variant (diff) in this reftile
 755                                         // region, whether a given genome
 756                                         // phase/allele (1) has the variant, (0) has
 757                                         // =ref or a different variant in that
 758                                         // position, or (-1) is lacking
 759                                         // coverage / couldn't be diffed.
 760                                         hgvsCol := hgvsColSet{}
 761                                         for _, diffs := range variantDiffs {
 762                                                 for _, diff := range diffs {
 763                                                         if _, ok := hgvsCol[diff]; ok {
 764                                                                 continue
 765                                                         }
 766                                                         hgvsCol[diff] = [2][]int8{
 767                                                                 make([]int8, len(cmd.cgnames)),
 768                                                                 make([]int8, len(cmd.cgnames)),
 769                                                         }
 770                                                 }
 771                                         }
 772                                         for row, name := range cmd.cgnames {
 773                                                 variants := cgs[name].Variants[(tag-tagstart)*2:]
 774                                                 for ph := 0; ph < 2; ph++ {
 775                                                         v := variants[ph]
 776                                                         if int(v) >= len(remap) {
 777                                                                 v = 0
 778                                                         } else {
 779                                                                 v = remap[v]
 780                                                         }
 781                                                         if v == rt.variant {
 782                                                                 // hgvsCol[*][ph][row] is already 0
 783                                                         } else if len(variantDiffs[v]) == 0 {
 784                                                                 // lacking coverage / couldn't be diffed
 785                                                                 for _, col := range hgvsCol {
 786                                                                         col[ph][row] = -1
 787                                                                 }
 788                                                         } else {
 789                                                                 for _, diff := range variantDiffs[v] {
 790                                                                         hgvsCol[diff][ph][row] = 1
 791                                                                 }
 792                                                         }
 793                                                 }
 794                                         }
 795                                         for diff, colpair := range hgvsCol {
 796                                                 allele2homhet(colpair)
 797                                                 if !cmd.filterHGVScolpair(colpair) {
 798                                                         delete(hgvsCol, diff)
 799                                                 }
 800                                         }
 801                                         if len(hgvsCol) > 0 {
 802                                                 encodeHGVSTodo[rt.seqname] <- hgvsCol
 803                                         }
 804                                 }
 805                                 outcol++
 806                         }
 807                         err = annow.Flush()
 808                         if err != nil {
 809                                 return err
 810                         }
 811                         err = annof.Close()
 812                         if err != nil {
 813                                 return err
 814                         }
 815
 816                         if *onehotChunked {
 817                                 // transpose onehotChunk[col][row] to numpy[row*ncols+col]
 818                                 rows := len(cmd.cgnames)
 819                                 cols := len(onehotChunk)
 820                                 log.Infof("%04d: preparing onehot numpy (rows=%d, cols=%d, mem=%d)", infileIdx, rows, cols, rows*cols)
 821                                 throttleNumpyMem.Acquire()
 822                                 out := onehotcols2int8(onehotChunk)
 823                                 fnm := fmt.Sprintf("%s/onehot.%04d.npy", *outputDir, infileIdx)
 824                                 err = writeNumpyInt8(fnm, out, rows, cols)
 825                                 if err != nil {
 826                                         return err
 827                                 }
 828                                 fnm = fmt.Sprintf("%s/onehot-columns.%04d.npy", *outputDir, infileIdx)
 829                                 err = writeNumpyInt32(fnm, onehotXref2int32(onehotXref), 4, len(onehotXref))
 830                                 if err != nil {
 831                                         return err
 832                                 }
 833                                 debug.FreeOSMemory()
 834                                 throttleNumpyMem.Release()
 835                         }
 836                         if *onehotSingle || *onlyPCA {
 837                                 onehotIndirect[infileIdx] = onehotChunk2Indirect(onehotChunk)
 838                                 onehotChunkSize[infileIdx] = uint32(len(onehotChunk))
 839                                 onehotXrefs[infileIdx] = onehotXref
 840                                 n := len(onehotIndirect[infileIdx][0])
 841                                 log.Infof("%04d: keeping onehot coordinates in memory (n=%d, mem=%d)", infileIdx, n, n*8*2)
 842                         }
 843                         if !(*onehotSingle || *onehotChunked || *onlyPCA) || *mergeOutput || *hgvsSingle {
 844                                 log.Infof("%04d: preparing numpy (rows=%d, cols=%d)", infileIdx, len(cmd.cgnames), 2*outcol)
 845                                 throttleNumpyMem.Acquire()
 846                                 rows := len(cmd.cgnames)
 847                                 cols := 2 * outcol
 848                                 out := make([]int16, rows*cols)
 849                                 for row, name := range cmd.cgnames {
 850                                         outidx := row * cols
 851                                         for col, v := range cgs[name].Variants {
 852                                                 tag := tagstart + tagID(col/2)
 853                                                 if cmd.filter.MaxTag >= 0 && tag > tagID(cmd.filter.MaxTag) {
 854                                                         break
 855                                                 }
 856                                                 if rt := reftile[tag]; rt == nil || rt.excluded {
 857                                                         continue
 858                                                 }
 859                                                 if v == 0 {
 860                                                         out[outidx] = 0 // tag not found / spanning tile
 861                                                 } else if variants, ok := seq[tag]; ok && int(v) < len(variants) && len(variants[v].Sequence) > 0 {
 862                                                         out[outidx] = int16(variantRemap[tag-tagstart][v])
 863                                                 } else {
 864                                                         out[outidx] = -1 // low quality tile variant
 865                                                 }
 866                                                 if tag == cmd.debugTag {
 867                                                         log.Printf("tag %d row %d col %d outidx %d v %d out %d", tag, row, col, outidx, v, out[outidx])
 868                                                 }
 869                                                 outidx++
 870                                         }
 871                                 }
 872                                 seq = nil
 873                                 cgs = nil
 874                                 debug.FreeOSMemory()
 875                                 throttleNumpyMem.Release()
 876                                 if *mergeOutput || *hgvsSingle {
 877                                         log.Infof("%04d: matrix fragment %d rows x %d cols", infileIdx, rows, cols)
 878                                         toMerge[infileIdx] = out
 879                                 }
 880                                 if !*mergeOutput && !*onehotChunked && !*onehotSingle {
 881                                         fnm := fmt.Sprintf("%s/matrix.%04d.npy", *outputDir, infileIdx)
 882                                         err = writeNumpyInt16(fnm, out, rows, cols)
 883                                         if err != nil {
 884                                                 return err
 885                                         }
 886                                 }
 887                         }
 888                         debug.FreeOSMemory()
 889                         log.Infof("%s: done (%d/%d)", infile, int(atomic.AddInt64(&done, 1)), len(infiles))
 890                         return nil
 891                 })
 892         }
 893         if err = throttleMem.Wait(); err != nil {
 894                 return err
 895         }
 896
 897         if *hgvsChunked {
 898                 log.Info("flushing hgvsCols temp files")
 899                 for seqname := range refseq {
 900                         close(encodeHGVSTodo[seqname])
 901                 }
 902                 err = encodeHGVS.Wait()
 903                 if err != nil {
 904                         return err
 905                 }
 906                 for seqname := range refseq {
 907                         log.Infof("%s: reading hgvsCols from temp file", seqname)
 908                         f := tmpHGVSCols[seqname]
 909                         _, err = f.Seek(0, io.SeekStart)
 910                         if err != nil {
 911                                 return err
 912                         }
 913                         var hgvsCols hgvsColSet
 914                         dec := gob.NewDecoder(bufio.NewReaderSize(f, 1<<24))
 915                         for err == nil {
 916                                 err = dec.Decode(&hgvsCols)
 917                         }
 918                         if err != io.EOF {
 919                                 return err
 920                         }
 921                         log.Infof("%s: sorting %d hgvs variants", seqname, len(hgvsCols))
 922                         variants := make([]hgvs.Variant, 0, len(hgvsCols))
 923                         for v := range hgvsCols {
 924                                 variants = append(variants, v)
 925                         }
 926                         sort.Slice(variants, func(i, j int) bool {
 927                                 vi, vj := &variants[i], &variants[j]
 928                                 if vi.Position != vj.Position {
 929                                         return vi.Position < vj.Position
 930                                 } else if vi.Ref != vj.Ref {
 931                                         return vi.Ref < vj.Ref
 932                                 } else {
 933                                         return vi.New < vj.New
 934                                 }
 935                         })
 936                         rows := len(cmd.cgnames)
 937                         cols := len(variants) * 2
 938                         log.Infof("%s: building hgvs matrix (rows=%d, cols=%d, mem=%d)", seqname, rows, cols, rows*cols)
 939                         out := make([]int8, rows*cols)
 940                         for varIdx, variant := range variants {
 941                                 hgvsCols := hgvsCols[variant]
 942                                 for row := range cmd.cgnames {
 943                                         for ph := 0; ph < 2; ph++ {
 944                                                 out[row*cols+varIdx+ph] = hgvsCols[ph][row]
 945                                         }
 946                                 }
 947                         }
 948                         err = writeNumpyInt8(fmt.Sprintf("%s/hgvs.%s.npy", *outputDir, seqname), out, rows, cols)
 949                         if err != nil {
 950                                 return err
 951                         }
 952                         out = nil
 953
 954                         fnm := fmt.Sprintf("%s/hgvs.%s.annotations.csv", *outputDir, seqname)
 955                         log.Infof("%s: writing hgvs column labels to %s", seqname, fnm)
 956                         var hgvsLabels bytes.Buffer
 957                         for varIdx, variant := range variants {
 958                                 fmt.Fprintf(&hgvsLabels, "%d,%s:g.%s\n", varIdx, seqname, variant.String())
 959                         }
 960                         err = ioutil.WriteFile(fnm, hgvsLabels.Bytes(), 0666)
 961                         if err != nil {
 962                                 return err
 963                         }
 964                 }
 965         }
 966
 967         if *mergeOutput || *hgvsSingle {
 968                 var annow *bufio.Writer
 969                 var annof *os.File
 970                 if *mergeOutput {
 971                         annoFilename := fmt.Sprintf("%s/matrix.annotations.csv", *outputDir)
 972                         annof, err = os.Create(annoFilename)
 973                         if err != nil {
 974                                 return err
 975                         }
 976                         annow = bufio.NewWriterSize(annof, 1<<20)
 977                 }
 978
 979                 rows := len(cmd.cgnames)
 980                 cols := 0
 981                 for _, chunk := range toMerge {
 982                         cols += len(chunk) / rows
 983                 }
 984                 log.Infof("merging output matrix (rows=%d, cols=%d, mem=%d) and annotations", rows, cols, rows*cols*2)
 985                 var out []int16
 986                 if *mergeOutput {
 987                         out = make([]int16, rows*cols)
 988                 }
 989                 hgvsCols := map[string][2][]int16{} // hgvs -> [[g0,g1,g2,...], [g0,g1,g2,...]] (slice of genomes for each phase)
 990                 startcol := 0
 991                 for outIdx, chunk := range toMerge {
 992                         chunkcols := len(chunk) / rows
 993                         if *mergeOutput {
 994                                 for row := 0; row < rows; row++ {
 995                                         copy(out[row*cols+startcol:], chunk[row*chunkcols:(row+1)*chunkcols])
 996                                 }
 997                         }
 998                         toMerge[outIdx] = nil
 999
1000                         annotationsFilename := fmt.Sprintf("%s/matrix.%04d.annotations.csv", *outputDir, outIdx)
1001                         log.Infof("reading %s", annotationsFilename)
1002                         buf, err := os.ReadFile(annotationsFilename)
1003                         if err != nil {
1004                                 return err
1005                         }
1006                         if *mergeOutput {
1007                                 err = os.Remove(annotationsFilename)
1008                                 if err != nil {
1009                                         return err
1010                                 }
1011                         }
1012                         for _, line := range bytes.Split(buf, []byte{'\n'}) {
1013                                 if len(line) == 0 {
1014                                         continue
1015                                 }
1016                                 fields := bytes.SplitN(line, []byte{','}, 9)
1017                                 tag, _ := strconv.Atoi(string(fields[0]))
1018                                 incol, _ := strconv.Atoi(string(fields[1]))
1019                                 tileVariant, _ := strconv.Atoi(string(fields[2]))
1020                                 hgvsID := string(fields[3])
1021                                 seqname := string(fields[4])
1022                                 pos, _ := strconv.Atoi(string(fields[5]))
1023                                 refseq := fields[6]
1024                                 if hgvsID == "" {
1025                                         // Null entry for un-diffable
1026                                         // tile variant
1027                                         continue
1028                                 }
1029                                 if hgvsID == "=" {
1030                                         // Null entry for ref tile
1031                                         continue
1032                                 }
1033                                 if mask != nil && !mask.Check(strings.TrimPrefix(seqname, "chr"), pos, pos+len(refseq)) {
1034                                         // The tile intersects one of
1035                                         // the selected regions, but
1036                                         // this particular HGVS
1037                                         // variant does not.
1038                                         continue
1039                                 }
1040                                 hgvsColPair := hgvsCols[hgvsID]
1041                                 if hgvsColPair[0] == nil {
1042                                         // values in new columns start
1043                                         // out as -1 ("no data yet")
1044                                         // or 0 ("=ref") here, may
1045                                         // change to 1 ("hgvs variant
1046                                         // present") below, either on
1047                                         // this line or a future line.
1048                                         hgvsColPair = [2][]int16{make([]int16, len(cmd.cgnames)), make([]int16, len(cmd.cgnames))}
1049                                         rt, ok := reftile[tagID(tag)]
1050                                         if !ok {
1051                                                 err = fmt.Errorf("bug: seeing annotations for tag %d, but it has no reftile entry", tag)
1052                                                 return err
1053                                         }
1054                                         for ph := 0; ph < 2; ph++ {
1055                                                 for row := 0; row < rows; row++ {
1056                                                         v := chunk[row*chunkcols+incol*2+ph]
1057                                                         if tileVariantID(v) == rt.variant {
1058                                                                 hgvsColPair[ph][row] = 0
1059                                                         } else {
1060                                                                 hgvsColPair[ph][row] = -1
1061                                                         }
1062                                                 }
1063                                         }
1064                                         hgvsCols[hgvsID] = hgvsColPair
1065                                         if annow != nil {
1066                                                 hgvsref := hgvs.Variant{
1067                                                         Position: pos,
1068                                                         Ref:      string(refseq),
1069                                                         New:      string(refseq),
1070                                                 }
1071                                                 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])
1072                                         }
1073                                 }
1074                                 if annow != nil {
1075                                         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])
1076                                 }
1077                                 for ph := 0; ph < 2; ph++ {
1078                                         for row := 0; row < rows; row++ {
1079                                                 v := chunk[row*chunkcols+incol*2+ph]
1080                                                 if int(v) == tileVariant {
1081                                                         hgvsColPair[ph][row] = 1
1082                                                 }
1083                                         }
1084                                 }
1085                         }
1086
1087                         startcol += chunkcols
1088                 }
1089                 if *mergeOutput {
1090                         err = annow.Flush()
1091                         if err != nil {
1092                                 return err
1093                         }
1094                         err = annof.Close()
1095                         if err != nil {
1096                                 return err
1097                         }
1098                         err = writeNumpyInt16(fmt.Sprintf("%s/matrix.npy", *outputDir), out, rows, cols)
1099                         if err != nil {
1100                                 return err
1101                         }
1102                 }
1103                 out = nil
1104
1105                 if *hgvsSingle {
1106                         cols = len(hgvsCols) * 2
1107                         log.Printf("building hgvs-based matrix: %d rows x %d cols", rows, cols)
1108                         out = make([]int16, rows*cols)
1109                         hgvsIDs := make([]string, 0, cols/2)
1110                         for hgvsID := range hgvsCols {
1111                                 hgvsIDs = append(hgvsIDs, hgvsID)
1112                         }
1113                         sort.Strings(hgvsIDs)
1114                         var hgvsLabels bytes.Buffer
1115                         for idx, hgvsID := range hgvsIDs {
1116                                 fmt.Fprintf(&hgvsLabels, "%d,%s\n", idx, hgvsID)
1117                                 for ph := 0; ph < 2; ph++ {
1118                                         hgvscol := hgvsCols[hgvsID][ph]
1119                                         for row, val := range hgvscol {
1120                                                 out[row*cols+idx*2+ph] = val
1121                                         }
1122                                 }
1123                         }
1124                         err = writeNumpyInt16(fmt.Sprintf("%s/hgvs.npy", *outputDir), out, rows, cols)
1125                         if err != nil {
1126                                 return err
1127                         }
1128
1129                         fnm := fmt.Sprintf("%s/hgvs.annotations.csv", *outputDir)
1130                         log.Printf("writing hgvs labels: %s", fnm)
1131                         err = ioutil.WriteFile(fnm, hgvsLabels.Bytes(), 0777)
1132                         if err != nil {
1133                                 return err
1134                         }
1135                 }
1136         }
1137         if *onehotSingle || *onlyPCA {
1138                 nzCount := 0
1139                 for _, part := range onehotIndirect {
1140                         nzCount += len(part[0])
1141                 }
1142                 onehot := make([]uint32, nzCount*2) // [r,r,r,...,c,c,c,...]
1143                 var xrefs []onehotXref
1144                 chunkOffset := uint32(0)
1145                 outcol := 0
1146                 for i, part := range onehotIndirect {
1147                         for i := range part[1] {
1148                                 part[1][i] += chunkOffset
1149                         }
1150                         copy(onehot[outcol:], part[0])
1151                         copy(onehot[outcol+nzCount:], part[1])
1152                         xrefs = append(xrefs, onehotXrefs[i]...)
1153
1154                         outcol += len(part[0])
1155                         chunkOffset += onehotChunkSize[i]
1156
1157                         part[0] = nil
1158                         part[1] = nil
1159                         onehotXrefs[i] = nil
1160                         debug.FreeOSMemory()
1161                 }
1162                 if *onehotSingle {
1163                         fnm := fmt.Sprintf("%s/onehot.npy", *outputDir)
1164                         err = writeNumpyUint32(fnm, onehot, 2, nzCount)
1165                         if err != nil {
1166                                 return err
1167                         }
1168                         fnm = fmt.Sprintf("%s/onehot-columns.npy", *outputDir)
1169                         err = writeNumpyInt32(fnm, onehotXref2int32(xrefs), 5, len(xrefs))
1170                         if err != nil {
1171                                 return err
1172                         }
1173                 }
1174                 if *onlyPCA {
1175                         cols := 0
1176                         for _, c := range onehot[nzCount:] {
1177                                 if int(c) >= cols {
1178                                         cols = int(c) + 1
1179                                 }
1180                         }
1181                         if cols == 0 {
1182                                 return fmt.Errorf("cannot do PCA: one-hot matrix is empty")
1183                         }
1184                         log.Printf("have %d one-hot cols", cols)
1185                         stride := 1
1186                         for *maxPCATiles > 0 && cols > *maxPCATiles*2 {
1187                                 cols = (cols + 1) / 2
1188                                 stride = stride * 2
1189                         }
1190                         log.Printf("creating full matrix (%d rows) and training matrix (%d rows) with %d cols, stride %d", len(cmd.cgnames), cmd.trainingSetSize, cols, stride)
1191                         mtxFull := mat.NewDense(len(cmd.cgnames), cols, nil)
1192                         mtxTrain := mat.NewDense(cmd.trainingSetSize, cols, nil)
1193                         for i, c := range onehot[nzCount:] {
1194                                 if int(c/2)%stride == 0 {
1195                                         outcol := int(c/2)/stride*2 + int(c)%2
1196                                         mtxFull.Set(int(onehot[i]), outcol, 1)
1197                                         if trainRow := cmd.trainingSet[int(onehot[i])]; trainRow >= 0 {
1198                                                 mtxTrain.Set(trainRow, outcol, 1)
1199                                         }
1200                                 }
1201                         }
1202                         log.Print("fitting")
1203                         transformer := nlp.NewPCA(*pcaComponents)
1204                         transformer.Fit(mtxTrain.T())
1205                         log.Printf("transforming")
1206                         pca, err := transformer.Transform(mtxFull.T())
1207                         if err != nil {
1208                                 return err
1209                         }
1210                         pca = pca.T()
1211                         outrows, outcols := pca.Dims()
1212                         log.Printf("copying result to numpy output array: %d rows, %d cols", outrows, outcols)
1213                         out := make([]float64, outrows*outcols)
1214                         for i := 0; i < outrows; i++ {
1215                                 for j := 0; j < outcols; j++ {
1216                                         out[i*outcols+j] = pca.At(i, j)
1217                                 }
1218                         }
1219                         fnm := fmt.Sprintf("%s/pca.npy", *outputDir)
1220                         log.Printf("writing numpy: %s", fnm)
1221                         output, err := os.OpenFile(fnm, os.O_CREATE|os.O_TRUNC|os.O_WRONLY, 0777)
1222                         if err != nil {
1223                                 return err
1224                         }
1225                         npw, err := gonpy.NewWriter(nopCloser{output})
1226                         if err != nil {
1227                                 return fmt.Errorf("gonpy.NewWriter: %w", err)
1228                         }
1229                         npw.Shape = []int{outrows, outcols}
1230                         err = npw.WriteFloat64(out)
1231                         if err != nil {
1232                                 return fmt.Errorf("WriteFloat64: %w", err)
1233                         }
1234                         err = output.Close()
1235                         if err != nil {
1236                                 return err
1237                         }
1238                         log.Print("done")
1239
1240                         samplesOutFilename := *outputDir + "/samples.csv"
1241                         log.Infof("writing sample metadata to %s", samplesOutFilename)
1242                         var f *os.File
1243                         f, err = os.Create(samplesOutFilename)
1244                         if err != nil {
1245                                 return err
1246                         }
1247                         defer f.Close()
1248                         for i, si := range cmd.samples {
1249                                 var cc, tv string
1250                                 if si.isCase {
1251                                         cc = "1"
1252                                 } else if si.isControl {
1253                                         cc = "0"
1254                                 }
1255                                 if si.isTraining {
1256                                         tv = "1"
1257                                 } else {
1258                                         tv = "0"
1259                                 }
1260                                 var pcavals string
1261                                 for c := 0; c < outcols; c++ {
1262                                         pcavals += fmt.Sprintf(",%f", pca.At(i, c))
1263                                 }
1264                                 _, err = fmt.Fprintf(f, "%d,%s,%s,%s%s\n", i, si.id, cc, tv, pcavals)
1265                                 if err != nil {
1266                                         err = fmt.Errorf("write %s: %w", samplesOutFilename, err)
1267                                         return err
1268                                 }
1269                         }
1270                         err = f.Close()
1271                         if err != nil {
1272                                 err = fmt.Errorf("close %s: %w", samplesOutFilename, err)
1273                                 return err
1274                         }
1275                         log.Print("done")
1276                 }
1277         }
1278         if !*mergeOutput && !*onehotChunked && !*onehotSingle && !*onlyPCA {
1279                 tagoffsetFilename := *outputDir + "/chunk-tag-offset.csv"
1280                 log.Infof("writing tag offsets to %s", tagoffsetFilename)
1281                 var f *os.File
1282                 f, err = os.Create(tagoffsetFilename)
1283                 if err != nil {
1284                         return err
1285                 }
1286                 defer f.Close()
1287                 for idx, offset := range chunkStartTag {
1288                         _, err = fmt.Fprintf(f, "%q,%d\n", fmt.Sprintf("matrix.%04d.npy", idx), offset)
1289                         if err != nil {
1290                                 err = fmt.Errorf("write %s: %w", tagoffsetFilename, err)
1291                                 return err
1292                         }
1293                 }
1294                 err = f.Close()
1295                 if err != nil {
1296                         err = fmt.Errorf("close %s: %w", tagoffsetFilename, err)
1297                         return err
1298                 }
1299         }
1300
1301         return nil
1302 }
1303
1304 type sampleInfo struct {
1305         id            string
1306         isCase        bool
1307         isControl     bool
1308         isTraining    bool
1309         isValidation  bool
1310         pcaComponents []float64
1311 }
1312
1313 // Read samples.csv file with case/control and training/validation
1314 // flags.
1315 func (cmd *sliceNumpy) loadSampleInfo(samplesFilename string) ([]sampleInfo, error) {
1316         var si []sampleInfo
1317         f, err := open(samplesFilename)
1318         if err != nil {
1319                 return nil, err
1320         }
1321         buf, err := io.ReadAll(f)
1322         f.Close()
1323         if err != nil {
1324                 return nil, err
1325         }
1326         lineNum := 0
1327         for _, csv := range bytes.Split(buf, []byte{'\n'}) {
1328                 lineNum++
1329                 if len(csv) == 0 {
1330                         continue
1331                 }
1332                 split := strings.Split(string(csv), ",")
1333                 if len(split) != 4 {
1334                         return nil, fmt.Errorf("%d fields != 4 in %s line %d: %q", len(split), samplesFilename, lineNum, csv)
1335                 }
1336                 if split[0] == "Index" && split[1] == "SampleID" && split[2] == "CaseControl" && split[3] == "TrainingValidation" {
1337                         continue
1338                 }
1339                 idx, err := strconv.Atoi(split[0])
1340                 if err != nil {
1341                         if lineNum == 1 {
1342                                 return nil, fmt.Errorf("header does not look right: %q", csv)
1343                         }
1344                         return nil, fmt.Errorf("%s line %d: index: %s", samplesFilename, lineNum, err)
1345                 }
1346                 if idx != len(si) {
1347                         return nil, fmt.Errorf("%s line %d: index %d out of order", samplesFilename, lineNum, idx)
1348                 }
1349                 si = append(si, sampleInfo{
1350                         id:           split[1],
1351                         isCase:       split[2] == "1",
1352                         isControl:    split[2] == "0",
1353                         isTraining:   split[3] == "1",
1354                         isValidation: split[3] == "0",
1355                 })
1356         }
1357         return si, nil
1358 }
1359
1360 func (cmd *sliceNumpy) filterHGVScolpair(colpair [2][]int8) bool {
1361         if cmd.chi2PValue >= 1 {
1362                 return true
1363         }
1364         col0 := make([]bool, 0, len(cmd.chi2Cases))
1365         col1 := make([]bool, 0, len(cmd.chi2Cases))
1366         cases := make([]bool, 0, len(cmd.chi2Cases))
1367         for i, c := range cmd.chi2Cases {
1368                 if colpair[0][i] < 0 {
1369                         continue
1370                 }
1371                 col0 = append(col0, colpair[0][i] != 0)
1372                 col1 = append(col1, colpair[1][i] != 0)
1373                 cases = append(cases, c)
1374         }
1375         return len(cases) >= cmd.minCoverage &&
1376                 (pvalue(col0, cases) <= cmd.chi2PValue || pvalue(col1, cases) <= cmd.chi2PValue)
1377 }
1378
1379 func writeNumpyUint32(fnm string, out []uint32, rows, cols int) error {
1380         output, err := os.Create(fnm)
1381         if err != nil {
1382                 return err
1383         }
1384         defer output.Close()
1385         bufw := bufio.NewWriterSize(output, 1<<26)
1386         npw, err := gonpy.NewWriter(nopCloser{bufw})
1387         if err != nil {
1388                 return err
1389         }
1390         log.WithFields(log.Fields{
1391                 "filename": fnm,
1392                 "rows":     rows,
1393                 "cols":     cols,
1394                 "bytes":    rows * cols * 4,
1395         }).Infof("writing numpy: %s", fnm)
1396         npw.Shape = []int{rows, cols}
1397         npw.WriteUint32(out)
1398         err = bufw.Flush()
1399         if err != nil {
1400                 return err
1401         }
1402         return output.Close()
1403 }
1404
1405 func writeNumpyInt32(fnm string, out []int32, rows, cols int) error {
1406         output, err := os.Create(fnm)
1407         if err != nil {
1408                 return err
1409         }
1410         defer output.Close()
1411         bufw := bufio.NewWriterSize(output, 1<<26)
1412         npw, err := gonpy.NewWriter(nopCloser{bufw})
1413         if err != nil {
1414                 return err
1415         }
1416         log.WithFields(log.Fields{
1417                 "filename": fnm,
1418                 "rows":     rows,
1419                 "cols":     cols,
1420                 "bytes":    rows * cols * 4,
1421         }).Infof("writing numpy: %s", fnm)
1422         npw.Shape = []int{rows, cols}
1423         npw.WriteInt32(out)
1424         err = bufw.Flush()
1425         if err != nil {
1426                 return err
1427         }
1428         return output.Close()
1429 }
1430
1431 func writeNumpyInt16(fnm string, out []int16, rows, cols int) error {
1432         output, err := os.Create(fnm)
1433         if err != nil {
1434                 return err
1435         }
1436         defer output.Close()
1437         bufw := bufio.NewWriterSize(output, 1<<26)
1438         npw, err := gonpy.NewWriter(nopCloser{bufw})
1439         if err != nil {
1440                 return err
1441         }
1442         log.WithFields(log.Fields{
1443                 "filename": fnm,
1444                 "rows":     rows,
1445                 "cols":     cols,
1446                 "bytes":    rows * cols * 2,
1447         }).Infof("writing numpy: %s", fnm)
1448         npw.Shape = []int{rows, cols}
1449         npw.WriteInt16(out)
1450         err = bufw.Flush()
1451         if err != nil {
1452                 return err
1453         }
1454         return output.Close()
1455 }
1456
1457 func writeNumpyInt8(fnm string, out []int8, rows, cols int) error {
1458         output, err := os.Create(fnm)
1459         if err != nil {
1460                 return err
1461         }
1462         defer output.Close()
1463         bufw := bufio.NewWriterSize(output, 1<<26)
1464         npw, err := gonpy.NewWriter(nopCloser{bufw})
1465         if err != nil {
1466                 return err
1467         }
1468         log.WithFields(log.Fields{
1469                 "filename": fnm,
1470                 "rows":     rows,
1471                 "cols":     cols,
1472                 "bytes":    rows * cols,
1473         }).Infof("writing numpy: %s", fnm)
1474         npw.Shape = []int{rows, cols}
1475         npw.WriteInt8(out)
1476         err = bufw.Flush()
1477         if err != nil {
1478                 return err
1479         }
1480         return output.Close()
1481 }
1482
1483 func allele2homhet(colpair [2][]int8) {
1484         a, b := colpair[0], colpair[1]
1485         for i, av := range a {
1486                 bv := b[i]
1487                 if av < 0 || bv < 0 {
1488                         // no-call
1489                         a[i], b[i] = -1, -1
1490                 } else if av > 0 && bv > 0 {
1491                         // hom
1492                         a[i], b[i] = 1, 0
1493                 } else if av > 0 || bv > 0 {
1494                         // het
1495                         a[i], b[i] = 0, 1
1496                 } else {
1497                         // ref (or a different variant in same position)
1498                         // (this is a no-op) a[i], b[i] = 0, 0
1499                 }
1500         }
1501 }
1502
1503 type onehotXref struct {
1504         tag     tagID
1505         variant tileVariantID
1506         hom     bool
1507         pvalue  float64
1508 }
1509
1510 const onehotXrefSize = unsafe.Sizeof(onehotXref{})
1511
1512 // Build onehot matrix (m[tileVariantIndex][genome] == 0 or 1) for all
1513 // variants of a single tile/tag#.
1514 //
1515 // Return nil if no tile variant passes Χ² filter.
1516 func (cmd *sliceNumpy) tv2homhet(cgs map[string]CompactGenome, maxv tileVariantID, remap []tileVariantID, tag, chunkstarttag tagID, seq map[tagID][]TileVariant) ([][]int8, []onehotXref) {
1517         if tag == cmd.debugTag {
1518                 tv := make([]tileVariantID, len(cmd.cgnames)*2)
1519                 for i, name := range cmd.cgnames {
1520                         copy(tv[i*2:(i+1)*2], cgs[name].Variants[(tag-chunkstarttag)*2:])
1521                 }
1522                 log.WithFields(logrus.Fields{
1523                         "cgs[i].Variants[tag*2+j]": tv,
1524                         "maxv":                     maxv,
1525                         "remap":                    remap,
1526                         "tag":                      tag,
1527                         "chunkstarttag":            chunkstarttag,
1528                 }).Info("tv2homhet()")
1529         }
1530         if maxv < 1 || (maxv < 2 && !cmd.includeVariant1) {
1531                 // everyone has the most common variant (of the variants we don't drop)
1532                 return nil, nil
1533         }
1534         tagoffset := tag - chunkstarttag
1535         coverage := 0
1536         for _, cg := range cgs {
1537                 alleles := 0
1538                 for _, v := range cg.Variants[tagoffset*2 : tagoffset*2+2] {
1539                         if v > 0 && int(v) < len(seq[tag]) && len(seq[tag][v].Sequence) > 0 {
1540                                 alleles++
1541                         }
1542                 }
1543                 if alleles == 2 {
1544                         coverage++
1545                 }
1546         }
1547         if coverage < cmd.minCoverage {
1548                 return nil, nil
1549         }
1550         obs := make([][]bool, (maxv+1)*2) // 2 slices (hom + het) for each variant#
1551         for i := range obs {
1552                 obs[i] = make([]bool, cmd.trainingSetSize)
1553         }
1554         for cgid, name := range cmd.cgnames {
1555                 tsid := cmd.trainingSet[cgid]
1556                 if tsid < 0 {
1557                         continue
1558                 }
1559                 cgvars := cgs[name].Variants[tagoffset*2:]
1560                 tv0, tv1 := remap[cgvars[0]], remap[cgvars[1]]
1561                 for v := tileVariantID(1); v <= maxv; v++ {
1562                         if tv0 == v && tv1 == v {
1563                                 obs[v*2][tsid] = true
1564                         } else if tv0 == v || tv1 == v {
1565                                 obs[v*2+1][tsid] = true
1566                         }
1567                 }
1568         }
1569         var onehot [][]int8
1570         var xref []onehotXref
1571         for col := 2; col < len(obs); col++ {
1572                 // col 0,1 correspond to tile variant 0, i.e.,
1573                 // no-call; col 2,3 correspond to the most common
1574                 // variant; so we (normally) start at col 4.
1575                 if col < 4 && !cmd.includeVariant1 {
1576                         continue
1577                 }
1578                 p := pvalue(obs[col], cmd.chi2Cases)
1579                 if cmd.chi2PValue < 1 && !(p < cmd.chi2PValue) {
1580                         continue
1581                 }
1582                 onehot = append(onehot, bool2int8(obs[col]))
1583                 xref = append(xref, onehotXref{
1584                         tag:     tag,
1585                         variant: tileVariantID(col >> 1),
1586                         hom:     col&1 == 0,
1587                         pvalue:  p,
1588                 })
1589         }
1590         return onehot, xref
1591 }
1592
1593 func bool2int8(in []bool) []int8 {
1594         out := make([]int8, len(in))
1595         for i, v := range in {
1596                 if v {
1597                         out[i] = 1
1598                 }
1599         }
1600         return out
1601 }
1602
1603 // convert a []onehotXref with length N to a numpy-style []int32
1604 // matrix with N columns, one row per field of onehotXref struct.
1605 //
1606 // Hom/het row contains hom=0, het=1.
1607 //
1608 // P-value row contains 1000000x actual p-value.
1609 func onehotXref2int32(xrefs []onehotXref) []int32 {
1610         xcols := len(xrefs)
1611         xdata := make([]int32, 5*xcols)
1612         for i, xref := range xrefs {
1613                 xdata[i] = int32(xref.tag)
1614                 xdata[xcols+i] = int32(xref.variant)
1615                 if xref.hom {
1616                         xdata[xcols*2+i] = 1
1617                 }
1618                 xdata[xcols*3+i] = int32(xref.pvalue * 1000000)
1619                 xdata[xcols*4+i] = int32(-math.Log10(xref.pvalue) * 1000000)
1620         }
1621         return xdata
1622 }
1623
1624 // transpose onehot data from in[col][row] to numpy-style
1625 // out[row*cols+col].
1626 func onehotcols2int8(in [][]int8) []int8 {
1627         if len(in) == 0 {
1628                 return nil
1629         }
1630         cols := len(in)
1631         rows := len(in[0])
1632         out := make([]int8, rows*cols)
1633         for row := 0; row < rows; row++ {
1634                 outrow := out[row*cols:]
1635                 for col, incol := range in {
1636                         outrow[col] = incol[row]
1637                 }
1638         }
1639         return out
1640 }
1641
1642 // Return [2][]uint32{rowIndices, colIndices} indicating which
1643 // elements of matrixT[c][r] have non-zero values.
1644 func onehotChunk2Indirect(matrixT [][]int8) [2][]uint32 {
1645         var nz [2][]uint32
1646         for c, col := range matrixT {
1647                 for r, val := range col {
1648                         if val != 0 {
1649                                 nz[0] = append(nz[0], uint32(r))
1650                                 nz[1] = append(nz[1], uint32(c))
1651                         }
1652                 }
1653         }
1654         return nz
1655 }