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         pcaComponents   int
  47         minCoverage     int
  48         includeVariant1 bool
  49         debugTag        tagID
  50
  51         cgnames         []string
  52         samples         []sampleInfo
  53         trainingSet     []int // samples index => training set index, or -1 if not in training set
  54         trainingSetSize int
  55         pvalue          func(onehot []bool) float64
  56 }
  57
  58 func (cmd *sliceNumpy) RunCommand(prog string, args []string, stdin io.Reader, stdout, stderr io.Writer) int {
  59         err := cmd.run(prog, args, stdin, stdout, stderr)
  60         if err != nil {
  61                 fmt.Fprintf(stderr, "%s\n", err)
  62                 return 1
  63         }
  64         return 0
  65 }
  66
  67 func (cmd *sliceNumpy) run(prog string, args []string, stdin io.Reader, stdout, stderr io.Writer) error {
  68         flags := flag.NewFlagSet("", flag.ContinueOnError)
  69         flags.SetOutput(stderr)
  70         pprof := flags.String("pprof", "", "serve Go profile data at http://`[addr]:port`")
  71         runlocal := flags.Bool("local", false, "run on local host (default: run in an arvados container)")
  72         arvadosRAM := flags.Int("arvados-ram", 750000000000, "amount of memory to request for arvados container (`bytes`)")
  73         arvadosVCPUs := flags.Int("arvados-vcpus", 96, "number of VCPUs to request for arvados container")
  74         projectUUID := flags.String("project", "", "project `UUID` for output data")
  75         priority := flags.Int("priority", 500, "container request priority")
  76         inputDir := flags.String("input-dir", "./in", "input `directory`")
  77         outputDir := flags.String("output-dir", "./out", "output `directory`")
  78         ref := flags.String("ref", "", "reference name (if blank, choose last one that appears in input)")
  79         regionsFilename := flags.String("regions", "", "only output columns/annotations that intersect regions in specified bed `file`")
  80         expandRegions := flags.Int("expand-regions", 0, "expand specified regions by `N` base pairs on each side`")
  81         mergeOutput := flags.Bool("merge-output", false, "merge output into one matrix.npy and one matrix.annotations.csv")
  82         hgvsSingle := flags.Bool("single-hgvs-matrix", false, "also generate hgvs-based matrix")
  83         hgvsChunked := flags.Bool("chunked-hgvs-matrix", false, "also generate hgvs-based matrix per chromosome")
  84         onehotSingle := flags.Bool("single-onehot", false, "generate one-hot tile-based matrix")
  85         onehotChunked := flags.Bool("chunked-onehot", false, "generate one-hot tile-based matrix per input chunk")
  86         samplesFilename := flags.String("samples", "", "`samples.csv` file with training/validation and case/control groups (see 'lightning choose-samples')")
  87         caseControlOnly := flags.Bool("case-control-only", false, "drop samples that are not in case/control groups")
  88         onlyPCA := flags.Bool("pca", false, "run principal component analysis, write components to pca.npy and samples.csv")
  89         flags.IntVar(&cmd.pcaComponents, "pca-components", 4, "number of PCA components to compute / use in logistic regression")
  90         maxPCATiles := flags.Int("max-pca-tiles", 0, "maximum tiles to use as PCA input (filter, then drop every 2nd colum pair until below max)")
  91         debugTag := flags.Int("debug-tag", -1, "log debugging details about specified tag")
  92         flags.IntVar(&cmd.threads, "threads", 16, "number of memory-hungry assembly threads, and number of VCPUs to request for arvados container")
  93         flags.Float64Var(&cmd.chi2PValue, "chi2-p-value", 1, "do Χ² test (or logistic regression if -samples file has PCA components) and omit columns with p-value above this threshold")
  94         flags.BoolVar(&cmd.includeVariant1, "include-variant-1", false, "include most common variant when building one-hot matrix")
  95         cmd.filter.Flags(flags)
  96         err := flags.Parse(args)
  97         if err == flag.ErrHelp {
  98                 return nil
  99         } else if err != nil {
 100                 return err
 101         } else if flags.NArg() > 0 {
 102                 return fmt.Errorf("errant command line arguments after parsed flags: %v", flags.Args())
 103         }
 104
 105         if *pprof != "" {
 106                 go func() {
 107                         log.Println(http.ListenAndServe(*pprof, nil))
 108                 }()
 109         }
 110
 111         if cmd.chi2PValue != 1 && *samplesFilename == "" {
 112                 return fmt.Errorf("cannot use provided -chi2-p-value=%f because -samples= value is empty", cmd.chi2PValue)
 113         }
 114
 115         cmd.debugTag = tagID(*debugTag)
 116
 117         if !*runlocal {
 118                 runner := arvadosContainerRunner{
 119                         Name:        "lightning slice-numpy",
 120                         Client:      arvados.NewClientFromEnv(),
 121                         ProjectUUID: *projectUUID,
 122                         RAM:         int64(*arvadosRAM),
 123                         VCPUs:       *arvadosVCPUs,
 124                         Priority:    *priority,
 125                         KeepCache:   2,
 126                         APIAccess:   true,
 127                 }
 128                 err = runner.TranslatePaths(inputDir, regionsFilename, samplesFilename)
 129                 if err != nil {
 130                         return err
 131                 }
 132                 runner.Args = []string{"slice-numpy", "-local=true",
 133                         "-pprof=:6060",
 134                         "-input-dir=" + *inputDir,
 135                         "-output-dir=/mnt/output",
 136                         "-threads=" + fmt.Sprintf("%d", cmd.threads),
 137                         "-regions=" + *regionsFilename,
 138                         "-expand-regions=" + fmt.Sprintf("%d", *expandRegions),
 139                         "-merge-output=" + fmt.Sprintf("%v", *mergeOutput),
 140                         "-single-hgvs-matrix=" + fmt.Sprintf("%v", *hgvsSingle),
 141                         "-chunked-hgvs-matrix=" + fmt.Sprintf("%v", *hgvsChunked),
 142                         "-single-onehot=" + fmt.Sprintf("%v", *onehotSingle),
 143                         "-chunked-onehot=" + fmt.Sprintf("%v", *onehotChunked),
 144                         "-samples=" + *samplesFilename,
 145                         "-case-control-only=" + fmt.Sprintf("%v", *caseControlOnly),
 146                         "-pca=" + fmt.Sprintf("%v", *onlyPCA),
 147                         "-pca-components=" + fmt.Sprintf("%d", cmd.pcaComponents),
 148                         "-max-pca-tiles=" + fmt.Sprintf("%d", *maxPCATiles),
 149                         "-chi2-p-value=" + fmt.Sprintf("%f", cmd.chi2PValue),
 150                         "-include-variant-1=" + fmt.Sprintf("%v", cmd.includeVariant1),
 151                         "-debug-tag=" + fmt.Sprintf("%d", cmd.debugTag),
 152                 }
 153                 runner.Args = append(runner.Args, cmd.filter.Args()...)
 154                 var output string
 155                 output, err = runner.Run()
 156                 if err != nil {
 157                         return err
 158                 }
 159                 fmt.Fprintln(stdout, output)
 160                 return nil
 161         }
 162
 163         infiles, err := allFiles(*inputDir, matchGobFile)
 164         if err != nil {
 165                 return err
 166         }
 167         if len(infiles) == 0 {
 168                 err = fmt.Errorf("no input files found in %s", *inputDir)
 169                 return err
 170         }
 171         sort.Strings(infiles)
 172
 173         var refseq map[string][]tileLibRef
 174         var reftiledata = make(map[tileLibRef][]byte, 11000000)
 175         in0, err := open(infiles[0])
 176         if err != nil {
 177                 return err
 178         }
 179
 180         matchGenome, err := regexp.Compile(cmd.filter.MatchGenome)
 181         if err != nil {
 182                 err = fmt.Errorf("-match-genome: invalid regexp: %q", cmd.filter.MatchGenome)
 183                 return err
 184         }
 185
 186         if *samplesFilename != "" {
 187                 cmd.samples, err = loadSampleInfo(*samplesFilename)
 188                 if err != nil {
 189                         return err
 190                 }
 191                 if len(cmd.samples[0].pcaComponents) > 0 {
 192                         cmd.pvalue = glmPvalueFunc(cmd.samples, cmd.pcaComponents)
 193                 }
 194         } else if *caseControlOnly {
 195                 return fmt.Errorf("-case-control-only does not make sense without -samples")
 196         }
 197
 198         cmd.cgnames = nil
 199         var tagset [][]byte
 200         err = DecodeLibrary(in0, strings.HasSuffix(infiles[0], ".gz"), func(ent *LibraryEntry) error {
 201                 if len(ent.TagSet) > 0 {
 202                         tagset = ent.TagSet
 203                 }
 204                 for _, cseq := range ent.CompactSequences {
 205                         if cseq.Name == *ref || *ref == "" {
 206                                 refseq = cseq.TileSequences
 207                         }
 208                 }
 209                 for _, cg := range ent.CompactGenomes {
 210                         if matchGenome.MatchString(cg.Name) {
 211                                 cmd.cgnames = append(cmd.cgnames, cg.Name)
 212                         }
 213                 }
 214                 for _, tv := range ent.TileVariants {
 215                         if tv.Ref {
 216                                 reftiledata[tileLibRef{tv.Tag, tv.Variant}] = tv.Sequence
 217                         }
 218                 }
 219                 return nil
 220         })
 221         if err != nil {
 222                 return err
 223         }
 224         in0.Close()
 225         if refseq == nil {
 226                 err = fmt.Errorf("%s: reference sequence not found", infiles[0])
 227                 return err
 228         }
 229         if len(tagset) == 0 {
 230                 err = fmt.Errorf("tagset not found")
 231                 return err
 232         }
 233
 234         taglib := &tagLibrary{}
 235         err = taglib.setTags(tagset)
 236         if err != nil {
 237                 return err
 238         }
 239         taglen := taglib.TagLen()
 240         sort.Strings(cmd.cgnames)
 241
 242         if len(cmd.cgnames) == 0 {
 243                 return fmt.Errorf("fatal: 0 matching samples in library, nothing to do")
 244         }
 245         cmd.trainingSet = make([]int, len(cmd.cgnames))
 246         if *samplesFilename == "" {
 247                 cmd.trainingSetSize = len(cmd.cgnames)
 248                 for i, name := range cmd.cgnames {
 249                         cmd.samples = append(cmd.samples, sampleInfo{
 250                                 id:         trimFilenameForLabel(name),
 251                                 isTraining: true,
 252                         })
 253                         cmd.trainingSet[i] = i
 254                 }
 255         } else if len(cmd.cgnames) != len(cmd.samples) {
 256                 return fmt.Errorf("mismatched sample list: %d samples in library, %d in %s", len(cmd.cgnames), len(cmd.samples), *samplesFilename)
 257         } else {
 258                 for i, name := range cmd.cgnames {
 259                         if s := trimFilenameForLabel(name); s != cmd.samples[i].id {
 260                                 return fmt.Errorf("mismatched sample list: sample %d is %q in library, %q in %s", i, s, cmd.samples[i].id, *samplesFilename)
 261                         }
 262                 }
 263                 if *caseControlOnly {
 264                         for i := 0; i < len(cmd.samples); i++ {
 265                                 if !cmd.samples[i].isTraining && !cmd.samples[i].isValidation {
 266                                         if i+1 < len(cmd.samples) {
 267                                                 copy(cmd.samples[i:], cmd.samples[i+1:])
 268                                                 copy(cmd.cgnames[i:], cmd.cgnames[i+1:])
 269                                         }
 270                                         cmd.samples = cmd.samples[:len(cmd.samples)-1]
 271                                         cmd.cgnames = cmd.cgnames[:len(cmd.cgnames)-1]
 272                                         i--
 273                                 }
 274                         }
 275                 }
 276                 cmd.chi2Cases = nil
 277                 cmd.trainingSetSize = 0
 278                 for i := range cmd.cgnames {
 279                         if cmd.samples[i].isTraining {
 280                                 cmd.trainingSet[i] = cmd.trainingSetSize
 281                                 cmd.trainingSetSize++
 282                                 cmd.chi2Cases = append(cmd.chi2Cases, cmd.samples[i].isCase)
 283                         } else {
 284                                 cmd.trainingSet[i] = -1
 285                         }
 286                 }
 287                 cmd.pvalue = func(onehot []bool) float64 {
 288                         return pvalue(onehot, cmd.chi2Cases)
 289                 }
 290         }
 291         if cmd.filter.MinCoverage == 1 {
 292                 // In the generic formula below, floating point
 293                 // arithmetic can effectively push the coverage
 294                 // threshold above 1.0, which is impossible/useless.
 295                 // 1.0 needs to mean exactly 100% coverage.
 296                 cmd.minCoverage = len(cmd.cgnames)
 297         } else {
 298                 cmd.minCoverage = int(math.Ceil(cmd.filter.MinCoverage * float64(len(cmd.cgnames))))
 299         }
 300
 301         {
 302                 samplesOutFilename := *outputDir + "/samples.csv"
 303                 log.Infof("writing sample metadata to %s", samplesOutFilename)
 304                 var f *os.File
 305                 f, err = os.Create(samplesOutFilename)
 306                 if err != nil {
 307                         return err
 308                 }
 309                 defer f.Close()
 310                 for i, si := range cmd.samples {
 311                         var cc, tv string
 312                         if si.isCase {
 313                                 cc = "1"
 314                         } else if si.isControl {
 315                                 cc = "0"
 316                         }
 317                         if si.isTraining {
 318                                 tv = "1"
 319                         } else {
 320                                 tv = "0"
 321                         }
 322                         _, err = fmt.Fprintf(f, "%d,%s,%s,%s\n", i, si.id, cc, tv)
 323                         if err != nil {
 324                                 err = fmt.Errorf("write %s: %w", samplesOutFilename, err)
 325                                 return err
 326                         }
 327                 }
 328                 err = f.Close()
 329                 if err != nil {
 330                         err = fmt.Errorf("close %s: %w", samplesOutFilename, err)
 331                         return err
 332                 }
 333                 log.Print("done")
 334         }
 335
 336         log.Info("indexing reference tiles")
 337         type reftileinfo struct {
 338                 variant  tileVariantID
 339                 seqname  string // chr1
 340                 pos      int    // distance from start of chromosome to starttag
 341                 tiledata []byte // acgtggcaa...
 342                 excluded bool   // true if excluded by regions file
 343                 nexttag  tagID  // tagID of following tile (-1 for last tag of chromosome)
 344         }
 345         isdup := map[tagID]bool{}
 346         reftile := map[tagID]*reftileinfo{}
 347         for seqname, cseq := range refseq {
 348                 pos := 0
 349                 lastreftag := tagID(-1)
 350                 for _, libref := range cseq {
 351                         if cmd.filter.MaxTag >= 0 && libref.Tag > tagID(cmd.filter.MaxTag) {
 352                                 continue
 353                         }
 354                         tiledata := reftiledata[libref]
 355                         if len(tiledata) == 0 {
 356                                 err = fmt.Errorf("missing tiledata for tag %d variant %d in %s in ref", libref.Tag, libref.Variant, seqname)
 357                                 return err
 358                         }
 359                         foundthistag := false
 360                         taglib.FindAll(tiledata[:len(tiledata)-1], func(tagid tagID, offset, _ int) {
 361                                 if !foundthistag && tagid == libref.Tag {
 362                                         foundthistag = true
 363                                         return
 364                                 }
 365                                 if dupref, ok := reftile[tagid]; ok {
 366                                         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)
 367                                         delete(reftile, tagid)
 368                                 } else {
 369                                         log.Printf("found tag %d at offset %d inside tile variant %+v on %s @ %d", tagid, offset, libref, seqname, pos+offset+1)
 370                                 }
 371                                 isdup[tagid] = true
 372                         })
 373                         if isdup[libref.Tag] {
 374                                 log.Printf("dropping reference tile %+v from %s @ %d, tag not unique", libref, seqname, pos)
 375                         } else if reftile[libref.Tag] != nil {
 376                                 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)
 377                                 delete(reftile, libref.Tag)
 378                                 log.Printf("dropping reference tile %+v from %s @ %d, tag not unique", libref, seqname, pos)
 379                                 isdup[libref.Tag] = true
 380                         } else {
 381                                 reftile[libref.Tag] = &reftileinfo{
 382                                         seqname:  seqname,
 383                                         variant:  libref.Variant,
 384                                         tiledata: tiledata,
 385                                         pos:      pos,
 386                                         nexttag:  -1,
 387                                 }
 388                                 if lastreftag >= 0 {
 389                                         reftile[lastreftag].nexttag = libref.Tag
 390                                 }
 391                                 lastreftag = libref.Tag
 392                         }
 393                         pos += len(tiledata) - taglen
 394                 }
 395                 log.Printf("... %s done, len %d", seqname, pos+taglen)
 396         }
 397
 398         var mask *mask
 399         if *regionsFilename != "" {
 400                 log.Printf("loading regions from %s", *regionsFilename)
 401                 mask, err = makeMask(*regionsFilename, *expandRegions)
 402                 if err != nil {
 403                         return err
 404                 }
 405                 log.Printf("before applying mask, len(reftile) == %d", len(reftile))
 406                 log.Printf("deleting reftile entries for regions outside %d intervals", mask.Len())
 407                 for _, rt := range reftile {
 408                         if !mask.Check(strings.TrimPrefix(rt.seqname, "chr"), rt.pos, rt.pos+len(rt.tiledata)) {
 409                                 rt.excluded = true
 410                         }
 411                 }
 412                 log.Printf("after applying mask, len(reftile) == %d", len(reftile))
 413         }
 414
 415         type hgvsColSet map[hgvs.Variant][2][]int8
 416         encodeHGVS := throttle{Max: len(refseq)}
 417         encodeHGVSTodo := map[string]chan hgvsColSet{}
 418         tmpHGVSCols := map[string]*os.File{}
 419         if *hgvsChunked {
 420                 for seqname := range refseq {
 421                         var f *os.File
 422                         f, err = os.Create(*outputDir + "/tmp." + seqname + ".gob")
 423                         if err != nil {
 424                                 return err
 425                         }
 426                         defer os.Remove(f.Name())
 427                         bufw := bufio.NewWriterSize(f, 1<<24)
 428                         enc := gob.NewEncoder(bufw)
 429                         tmpHGVSCols[seqname] = f
 430                         todo := make(chan hgvsColSet, 128)
 431                         encodeHGVSTodo[seqname] = todo
 432                         encodeHGVS.Go(func() error {
 433                                 for colset := range todo {
 434                                         err := enc.Encode(colset)
 435                                         if err != nil {
 436                                                 encodeHGVS.Report(err)
 437                                                 for range todo {
 438                                                 }
 439                                                 return err
 440                                         }
 441                                 }
 442                                 return bufw.Flush()
 443                         })
 444                 }
 445         }
 446
 447         var toMerge [][]int16
 448         if *mergeOutput || *hgvsSingle {
 449                 toMerge = make([][]int16, len(infiles))
 450         }
 451         var onehotIndirect [][2][]uint32 // [chunkIndex][axis][index]
 452         var onehotChunkSize []uint32
 453         var onehotXrefs [][]onehotXref
 454         if *onehotSingle || *onlyPCA {
 455                 onehotIndirect = make([][2][]uint32, len(infiles))
 456                 onehotChunkSize = make([]uint32, len(infiles))
 457                 onehotXrefs = make([][]onehotXref, len(infiles))
 458         }
 459         chunkStartTag := make([]tagID, len(infiles))
 460
 461         throttleMem := throttle{Max: cmd.threads} // TODO: estimate using mem and data size
 462         throttleNumpyMem := throttle{Max: cmd.threads/2 + 1}
 463         log.Info("generating annotations and numpy matrix for each slice")
 464         var errSkip = errors.New("skip infile")
 465         var done int64
 466         for infileIdx, infile := range infiles {
 467                 infileIdx, infile := infileIdx, infile
 468                 throttleMem.Go(func() error {
 469                         seq := make(map[tagID][]TileVariant, 50000)
 470                         cgs := make(map[string]CompactGenome, len(cmd.cgnames))
 471                         f, err := open(infile)
 472                         if err != nil {
 473                                 return err
 474                         }
 475                         defer f.Close()
 476                         log.Infof("%04d: reading %s", infileIdx, infile)
 477                         err = DecodeLibrary(f, strings.HasSuffix(infile, ".gz"), func(ent *LibraryEntry) error {
 478                                 for _, tv := range ent.TileVariants {
 479                                         if tv.Ref {
 480                                                 continue
 481                                         }
 482                                         // Skip tile with no
 483                                         // corresponding ref tile, if
 484                                         // mask is in play (we can't
 485                                         // determine coordinates for
 486                                         // these)
 487                                         if mask != nil && reftile[tv.Tag] == nil {
 488                                                 continue
 489                                         }
 490                                         // Skip tile whose
 491                                         // corresponding ref tile is
 492                                         // outside target regions --
 493                                         // unless it's a potential
 494                                         // spanning tile.
 495                                         if mask != nil && reftile[tv.Tag].excluded &&
 496                                                 (int(tv.Tag+1) >= len(tagset) ||
 497                                                         (bytes.HasSuffix(tv.Sequence, tagset[tv.Tag+1]) && reftile[tv.Tag+1] != nil && !reftile[tv.Tag+1].excluded)) {
 498                                                 continue
 499                                         }
 500                                         if tv.Tag == cmd.debugTag {
 501                                                 log.Printf("infile %d %s tag %d variant %d hash %x", infileIdx, infile, tv.Tag, tv.Variant, tv.Blake2b[:3])
 502                                         }
 503                                         variants := seq[tv.Tag]
 504                                         if len(variants) == 0 {
 505                                                 variants = make([]TileVariant, 100)
 506                                         }
 507                                         for len(variants) <= int(tv.Variant) {
 508                                                 variants = append(variants, TileVariant{})
 509                                         }
 510                                         variants[int(tv.Variant)] = tv
 511                                         seq[tv.Tag] = variants
 512                                 }
 513                                 for _, cg := range ent.CompactGenomes {
 514                                         if cmd.filter.MaxTag >= 0 && cg.StartTag > tagID(cmd.filter.MaxTag) {
 515                                                 return errSkip
 516                                         }
 517                                         if !matchGenome.MatchString(cg.Name) {
 518                                                 continue
 519                                         }
 520                                         // pad to full slice size
 521                                         // to avoid out-of-bounds
 522                                         // checks later
 523                                         if sliceSize := 2 * int(cg.EndTag-cg.StartTag); len(cg.Variants) < sliceSize {
 524                                                 cg.Variants = append(cg.Variants, make([]tileVariantID, sliceSize-len(cg.Variants))...)
 525                                         }
 526                                         cgs[cg.Name] = cg
 527                                 }
 528                                 return nil
 529                         })
 530                         if err == errSkip {
 531                                 return nil
 532                         } else if err != nil {
 533                                 return fmt.Errorf("%04d: DecodeLibrary(%s): err", infileIdx, infile)
 534                         }
 535                         tagstart := cgs[cmd.cgnames[0]].StartTag
 536                         tagend := cgs[cmd.cgnames[0]].EndTag
 537                         chunkStartTag[infileIdx] = tagstart
 538
 539                         // TODO: filters
 540
 541                         log.Infof("%04d: renumber/dedup variants for tags %d-%d", infileIdx, tagstart, tagend)
 542                         variantRemap := make([][]tileVariantID, tagend-tagstart)
 543                         throttleCPU := throttle{Max: runtime.GOMAXPROCS(0)}
 544                         for tag, variants := range seq {
 545                                 tag, variants := tag, variants
 546                                 throttleCPU.Go(func() error {
 547                                         alleleCoverage := 0
 548                                         count := make(map[[blake2b.Size256]byte]int, len(variants))
 549
 550                                         rt := reftile[tag]
 551                                         if rt != nil {
 552                                                 count[blake2b.Sum256(rt.tiledata)] = 0
 553                                         }
 554
 555                                         for cgname, cg := range cgs {
 556                                                 idx := int(tag-tagstart) * 2
 557                                                 for allele := 0; allele < 2; allele++ {
 558                                                         v := cg.Variants[idx+allele]
 559                                                         if v > 0 && len(variants[v].Sequence) > 0 {
 560                                                                 count[variants[v].Blake2b]++
 561                                                                 alleleCoverage++
 562                                                         }
 563                                                         if v > 0 && tag == cmd.debugTag {
 564                                                                 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])
 565                                                         }
 566                                                 }
 567                                         }
 568                                         if alleleCoverage < cmd.minCoverage*2 {
 569                                                 idx := int(tag-tagstart) * 2
 570                                                 for _, cg := range cgs {
 571                                                         cg.Variants[idx] = 0
 572                                                         cg.Variants[idx+1] = 0
 573                                                 }
 574                                                 if tag == cmd.debugTag {
 575                                                         log.Printf("tag %d alleleCoverage %d < min %d, sample data wiped", tag, alleleCoverage, cmd.minCoverage*2)
 576                                                 }
 577                                                 return nil
 578                                         }
 579
 580                                         // hash[i] will be the hash of
 581                                         // the variant(s) that should
 582                                         // be at rank i (0-based).
 583                                         hash := make([][blake2b.Size256]byte, 0, len(count))
 584                                         for b := range count {
 585                                                 hash = append(hash, b)
 586                                         }
 587                                         sort.Slice(hash, func(i, j int) bool {
 588                                                 bi, bj := &hash[i], &hash[j]
 589                                                 if ci, cj := count[*bi], count[*bj]; ci != cj {
 590                                                         return ci > cj
 591                                                 } else {
 592                                                         return bytes.Compare((*bi)[:], (*bj)[:]) < 0
 593                                                 }
 594                                         })
 595                                         // rank[b] will be the 1-based
 596                                         // new variant number for
 597                                         // variants whose hash is b.
 598                                         rank := make(map[[blake2b.Size256]byte]tileVariantID, len(hash))
 599                                         for i, h := range hash {
 600                                                 rank[h] = tileVariantID(i + 1)
 601                                         }
 602                                         if tag == cmd.debugTag {
 603                                                 for h, r := range rank {
 604                                                         log.Printf("tag %d rank(%x) = %v", tag, h[:3], r)
 605                                                 }
 606                                         }
 607                                         // remap[v] will be the new
 608                                         // variant number for original
 609                                         // variant number v.
 610                                         remap := make([]tileVariantID, len(variants))
 611                                         for i, tv := range variants {
 612                                                 remap[i] = rank[tv.Blake2b]
 613                                         }
 614                                         if tag == cmd.debugTag {
 615                                                 for in, out := range remap {
 616                                                         if out > 0 {
 617                                                                 log.Printf("tag %d remap %d => %d", tag, in, out)
 618                                                         }
 619                                                 }
 620                                         }
 621                                         variantRemap[tag-tagstart] = remap
 622                                         if rt != nil {
 623                                                 refrank := rank[blake2b.Sum256(rt.tiledata)]
 624                                                 if tag == cmd.debugTag {
 625                                                         log.Printf("tag %d reftile variant %d => %d", tag, rt.variant, refrank)
 626                                                 }
 627                                                 rt.variant = refrank
 628                                         }
 629                                         return nil
 630                                 })
 631                         }
 632                         throttleCPU.Wait()
 633
 634                         var onehotChunk [][]int8
 635                         var onehotXref []onehotXref
 636
 637                         var annotationsFilename string
 638                         if *onlyPCA {
 639                                 annotationsFilename = "/dev/null"
 640                         } else {
 641                                 annotationsFilename = fmt.Sprintf("%s/matrix.%04d.annotations.csv", *outputDir, infileIdx)
 642                                 log.Infof("%04d: writing %s", infileIdx, annotationsFilename)
 643                         }
 644                         annof, err := os.Create(annotationsFilename)
 645                         if err != nil {
 646                                 return err
 647                         }
 648                         annow := bufio.NewWriterSize(annof, 1<<20)
 649                         outcol := 0
 650                         for tag := tagstart; tag < tagend; tag++ {
 651                                 rt := reftile[tag]
 652                                 if rt == nil && mask != nil {
 653                                         // With no ref tile, we don't
 654                                         // have coordinates to say
 655                                         // this is in the desired
 656                                         // regions -- so it's not.
 657                                         // TODO: handle ref spanning
 658                                         // tile case.
 659                                         continue
 660                                 }
 661                                 if rt != nil && rt.excluded {
 662                                         // TODO: don't skip yet --
 663                                         // first check for spanning
 664                                         // tile variants that
 665                                         // intersect non-excluded ref
 666                                         // tiles.
 667                                         continue
 668                                 }
 669                                 if cmd.filter.MaxTag >= 0 && tag > tagID(cmd.filter.MaxTag) {
 670                                         break
 671                                 }
 672                                 remap := variantRemap[tag-tagstart]
 673                                 if remap == nil {
 674                                         // was not assigned above,
 675                                         // because minCoverage
 676                                         outcol++
 677                                         continue
 678                                 }
 679                                 maxv := tileVariantID(0)
 680                                 for _, v := range remap {
 681                                         if maxv < v {
 682                                                 maxv = v
 683                                         }
 684                                 }
 685                                 if *onehotChunked || *onehotSingle || *onlyPCA {
 686                                         onehot, xrefs := cmd.tv2homhet(cgs, maxv, remap, tag, tagstart, seq)
 687                                         if tag == cmd.debugTag {
 688                                                 log.WithFields(logrus.Fields{
 689                                                         "onehot": onehot,
 690                                                         "xrefs":  xrefs,
 691                                                 }).Info("tv2homhet()")
 692                                         }
 693                                         onehotChunk = append(onehotChunk, onehot...)
 694                                         onehotXref = append(onehotXref, xrefs...)
 695                                 }
 696                                 if *onlyPCA {
 697                                         outcol++
 698                                         continue
 699                                 }
 700                                 if rt == nil {
 701                                         // Reference does not use any
 702                                         // variant of this tile
 703                                         //
 704                                         // TODO: diff against the
 705                                         // relevant portion of the
 706                                         // ref's spanning tile
 707                                         outcol++
 708                                         continue
 709                                 }
 710                                 fmt.Fprintf(annow, "%d,%d,%d,=,%s,%d,,,\n", tag, outcol, rt.variant, rt.seqname, rt.pos)
 711                                 variants := seq[tag]
 712                                 reftilestr := strings.ToUpper(string(rt.tiledata))
 713
 714                                 done := make([]bool, maxv+1)
 715                                 variantDiffs := make([][]hgvs.Variant, maxv+1)
 716                                 for v, tv := range variants {
 717                                         v := remap[v]
 718                                         if v == 0 || v == rt.variant || done[v] {
 719                                                 continue
 720                                         } else {
 721                                                 done[v] = true
 722                                         }
 723                                         if len(tv.Sequence) < taglen {
 724                                                 continue
 725                                         }
 726                                         // if reftilestr doesn't end
 727                                         // in the same tag as tv,
 728                                         // extend reftilestr with
 729                                         // following ref tiles until
 730                                         // it does (up to an arbitrary
 731                                         // sanity-check limit)
 732                                         reftilestr := reftilestr
 733                                         endtagstr := strings.ToUpper(string(tv.Sequence[len(tv.Sequence)-taglen:]))
 734                                         for i, rt := 0, rt; i < annotationMaxTileSpan && !strings.HasSuffix(reftilestr, endtagstr) && rt.nexttag >= 0; i++ {
 735                                                 rt = reftile[rt.nexttag]
 736                                                 if rt == nil {
 737                                                         break
 738                                                 }
 739                                                 reftilestr += strings.ToUpper(string(rt.tiledata[taglen:]))
 740                                         }
 741                                         if mask != nil && !mask.Check(strings.TrimPrefix(rt.seqname, "chr"), rt.pos, rt.pos+len(reftilestr)) {
 742                                                 continue
 743                                         }
 744                                         if !strings.HasSuffix(reftilestr, endtagstr) {
 745                                                 fmt.Fprintf(annow, "%d,%d,%d,,%s,%d,,,\n", tag, outcol, v, rt.seqname, rt.pos)
 746                                                 continue
 747                                         }
 748                                         if lendiff := len(reftilestr) - len(tv.Sequence); lendiff < -1000 || lendiff > 1000 {
 749                                                 fmt.Fprintf(annow, "%d,%d,%d,,%s,%d,,,\n", tag, outcol, v, rt.seqname, rt.pos)
 750                                                 continue
 751                                         }
 752                                         diffs, _ := hgvs.Diff(reftilestr, strings.ToUpper(string(tv.Sequence)), 0)
 753                                         for i := range diffs {
 754                                                 diffs[i].Position += rt.pos
 755                                         }
 756                                         for _, diff := range diffs {
 757                                                 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)
 758                                         }
 759                                         if *hgvsChunked {
 760                                                 variantDiffs[v] = diffs
 761                                         }
 762                                 }
 763                                 if *hgvsChunked {
 764                                         // We can now determine, for each HGVS
 765                                         // variant (diff) in this reftile
 766                                         // region, whether a given genome
 767                                         // phase/allele (1) has the variant, (0) has
 768                                         // =ref or a different variant in that
 769                                         // position, or (-1) is lacking
 770                                         // coverage / couldn't be diffed.
 771                                         hgvsCol := hgvsColSet{}
 772                                         for _, diffs := range variantDiffs {
 773                                                 for _, diff := range diffs {
 774                                                         if _, ok := hgvsCol[diff]; ok {
 775                                                                 continue
 776                                                         }
 777                                                         hgvsCol[diff] = [2][]int8{
 778                                                                 make([]int8, len(cmd.cgnames)),
 779                                                                 make([]int8, len(cmd.cgnames)),
 780                                                         }
 781                                                 }
 782                                         }
 783                                         for row, name := range cmd.cgnames {
 784                                                 variants := cgs[name].Variants[(tag-tagstart)*2:]
 785                                                 for ph := 0; ph < 2; ph++ {
 786                                                         v := variants[ph]
 787                                                         if int(v) >= len(remap) {
 788                                                                 v = 0
 789                                                         } else {
 790                                                                 v = remap[v]
 791                                                         }
 792                                                         if v == rt.variant {
 793                                                                 // hgvsCol[*][ph][row] is already 0
 794                                                         } else if len(variantDiffs[v]) == 0 {
 795                                                                 // lacking coverage / couldn't be diffed
 796                                                                 for _, col := range hgvsCol {
 797                                                                         col[ph][row] = -1
 798                                                                 }
 799                                                         } else {
 800                                                                 for _, diff := range variantDiffs[v] {
 801                                                                         hgvsCol[diff][ph][row] = 1
 802                                                                 }
 803                                                         }
 804                                                 }
 805                                         }
 806                                         for diff, colpair := range hgvsCol {
 807                                                 allele2homhet(colpair)
 808                                                 if !cmd.filterHGVScolpair(colpair) {
 809                                                         delete(hgvsCol, diff)
 810                                                 }
 811                                         }
 812                                         if len(hgvsCol) > 0 {
 813                                                 encodeHGVSTodo[rt.seqname] <- hgvsCol
 814                                         }
 815                                 }
 816                                 outcol++
 817                         }
 818                         err = annow.Flush()
 819                         if err != nil {
 820                                 return err
 821                         }
 822                         err = annof.Close()
 823                         if err != nil {
 824                                 return err
 825                         }
 826
 827                         if *onehotChunked {
 828                                 // transpose onehotChunk[col][row] to numpy[row*ncols+col]
 829                                 rows := len(cmd.cgnames)
 830                                 cols := len(onehotChunk)
 831                                 log.Infof("%04d: preparing onehot numpy (rows=%d, cols=%d, mem=%d)", infileIdx, rows, cols, rows*cols)
 832                                 throttleNumpyMem.Acquire()
 833                                 out := onehotcols2int8(onehotChunk)
 834                                 fnm := fmt.Sprintf("%s/onehot.%04d.npy", *outputDir, infileIdx)
 835                                 err = writeNumpyInt8(fnm, out, rows, cols)
 836                                 if err != nil {
 837                                         return err
 838                                 }
 839                                 fnm = fmt.Sprintf("%s/onehot-columns.%04d.npy", *outputDir, infileIdx)
 840                                 err = writeNumpyInt32(fnm, onehotXref2int32(onehotXref), 4, len(onehotXref))
 841                                 if err != nil {
 842                                         return err
 843                                 }
 844                                 debug.FreeOSMemory()
 845                                 throttleNumpyMem.Release()
 846                         }
 847                         if *onehotSingle || *onlyPCA {
 848                                 onehotIndirect[infileIdx] = onehotChunk2Indirect(onehotChunk)
 849                                 onehotChunkSize[infileIdx] = uint32(len(onehotChunk))
 850                                 onehotXrefs[infileIdx] = onehotXref
 851                                 n := len(onehotIndirect[infileIdx][0])
 852                                 log.Infof("%04d: keeping onehot coordinates in memory (n=%d, mem=%d)", infileIdx, n, n*8*2)
 853                         }
 854                         if !(*onehotSingle || *onehotChunked || *onlyPCA) || *mergeOutput || *hgvsSingle {
 855                                 log.Infof("%04d: preparing numpy (rows=%d, cols=%d)", infileIdx, len(cmd.cgnames), 2*outcol)
 856                                 throttleNumpyMem.Acquire()
 857                                 rows := len(cmd.cgnames)
 858                                 cols := 2 * outcol
 859                                 out := make([]int16, rows*cols)
 860                                 for row, name := range cmd.cgnames {
 861                                         outidx := row * cols
 862                                         for col, v := range cgs[name].Variants {
 863                                                 tag := tagstart + tagID(col/2)
 864                                                 if cmd.filter.MaxTag >= 0 && tag > tagID(cmd.filter.MaxTag) {
 865                                                         break
 866                                                 }
 867                                                 if rt := reftile[tag]; rt == nil || rt.excluded {
 868                                                         continue
 869                                                 }
 870                                                 if v == 0 {
 871                                                         out[outidx] = 0 // tag not found / spanning tile
 872                                                 } else if variants, ok := seq[tag]; ok && int(v) < len(variants) && len(variants[v].Sequence) > 0 {
 873                                                         out[outidx] = int16(variantRemap[tag-tagstart][v])
 874                                                 } else {
 875                                                         out[outidx] = -1 // low quality tile variant
 876                                                 }
 877                                                 if tag == cmd.debugTag {
 878                                                         log.Printf("tag %d row %d col %d outidx %d v %d out %d", tag, row, col, outidx, v, out[outidx])
 879                                                 }
 880                                                 outidx++
 881                                         }
 882                                 }
 883                                 seq = nil
 884                                 cgs = nil
 885                                 debug.FreeOSMemory()
 886                                 throttleNumpyMem.Release()
 887                                 if *mergeOutput || *hgvsSingle {
 888                                         log.Infof("%04d: matrix fragment %d rows x %d cols", infileIdx, rows, cols)
 889                                         toMerge[infileIdx] = out
 890                                 }
 891                                 if !*mergeOutput && !*onehotChunked && !*onehotSingle {
 892                                         fnm := fmt.Sprintf("%s/matrix.%04d.npy", *outputDir, infileIdx)
 893                                         err = writeNumpyInt16(fnm, out, rows, cols)
 894                                         if err != nil {
 895                                                 return err
 896                                         }
 897                                 }
 898                         }
 899                         debug.FreeOSMemory()
 900                         log.Infof("%s: done (%d/%d)", infile, int(atomic.AddInt64(&done, 1)), len(infiles))
 901                         return nil
 902                 })
 903         }
 904         if err = throttleMem.Wait(); err != nil {
 905                 return err
 906         }
 907
 908         if *hgvsChunked {
 909                 log.Info("flushing hgvsCols temp files")
 910                 for seqname := range refseq {
 911                         close(encodeHGVSTodo[seqname])
 912                 }
 913                 err = encodeHGVS.Wait()
 914                 if err != nil {
 915                         return err
 916                 }
 917                 for seqname := range refseq {
 918                         log.Infof("%s: reading hgvsCols from temp file", seqname)
 919                         f := tmpHGVSCols[seqname]
 920                         _, err = f.Seek(0, io.SeekStart)
 921                         if err != nil {
 922                                 return err
 923                         }
 924                         var hgvsCols hgvsColSet
 925                         dec := gob.NewDecoder(bufio.NewReaderSize(f, 1<<24))
 926                         for err == nil {
 927                                 err = dec.Decode(&hgvsCols)
 928                         }
 929                         if err != io.EOF {
 930                                 return err
 931                         }
 932                         log.Infof("%s: sorting %d hgvs variants", seqname, len(hgvsCols))
 933                         variants := make([]hgvs.Variant, 0, len(hgvsCols))
 934                         for v := range hgvsCols {
 935                                 variants = append(variants, v)
 936                         }
 937                         sort.Slice(variants, func(i, j int) bool {
 938                                 vi, vj := &variants[i], &variants[j]
 939                                 if vi.Position != vj.Position {
 940                                         return vi.Position < vj.Position
 941                                 } else if vi.Ref != vj.Ref {
 942                                         return vi.Ref < vj.Ref
 943                                 } else {
 944                                         return vi.New < vj.New
 945                                 }
 946                         })
 947                         rows := len(cmd.cgnames)
 948                         cols := len(variants) * 2
 949                         log.Infof("%s: building hgvs matrix (rows=%d, cols=%d, mem=%d)", seqname, rows, cols, rows*cols)
 950                         out := make([]int8, rows*cols)
 951                         for varIdx, variant := range variants {
 952                                 hgvsCols := hgvsCols[variant]
 953                                 for row := range cmd.cgnames {
 954                                         for ph := 0; ph < 2; ph++ {
 955                                                 out[row*cols+varIdx+ph] = hgvsCols[ph][row]
 956                                         }
 957                                 }
 958                         }
 959                         err = writeNumpyInt8(fmt.Sprintf("%s/hgvs.%s.npy", *outputDir, seqname), out, rows, cols)
 960                         if err != nil {
 961                                 return err
 962                         }
 963                         out = nil
 964
 965                         fnm := fmt.Sprintf("%s/hgvs.%s.annotations.csv", *outputDir, seqname)
 966                         log.Infof("%s: writing hgvs column labels to %s", seqname, fnm)
 967                         var hgvsLabels bytes.Buffer
 968                         for varIdx, variant := range variants {
 969                                 fmt.Fprintf(&hgvsLabels, "%d,%s:g.%s\n", varIdx, seqname, variant.String())
 970                         }
 971                         err = ioutil.WriteFile(fnm, hgvsLabels.Bytes(), 0666)
 972                         if err != nil {
 973                                 return err
 974                         }
 975                 }
 976         }
 977
 978         if *mergeOutput || *hgvsSingle {
 979                 var annow *bufio.Writer
 980                 var annof *os.File
 981                 if *mergeOutput {
 982                         annoFilename := fmt.Sprintf("%s/matrix.annotations.csv", *outputDir)
 983                         annof, err = os.Create(annoFilename)
 984                         if err != nil {
 985                                 return err
 986                         }
 987                         annow = bufio.NewWriterSize(annof, 1<<20)
 988                 }
 989
 990                 rows := len(cmd.cgnames)
 991                 cols := 0
 992                 for _, chunk := range toMerge {
 993                         cols += len(chunk) / rows
 994                 }
 995                 log.Infof("merging output matrix (rows=%d, cols=%d, mem=%d) and annotations", rows, cols, rows*cols*2)
 996                 var out []int16
 997                 if *mergeOutput {
 998                         out = make([]int16, rows*cols)
 999                 }
1000                 hgvsCols := map[string][2][]int16{} // hgvs -> [[g0,g1,g2,...], [g0,g1,g2,...]] (slice of genomes for each phase)
1001                 startcol := 0
1002                 for outIdx, chunk := range toMerge {
1003                         chunkcols := len(chunk) / rows
1004                         if *mergeOutput {
1005                                 for row := 0; row < rows; row++ {
1006                                         copy(out[row*cols+startcol:], chunk[row*chunkcols:(row+1)*chunkcols])
1007                                 }
1008                         }
1009                         toMerge[outIdx] = nil
1010
1011                         annotationsFilename := fmt.Sprintf("%s/matrix.%04d.annotations.csv", *outputDir, outIdx)
1012                         log.Infof("reading %s", annotationsFilename)
1013                         buf, err := os.ReadFile(annotationsFilename)
1014                         if err != nil {
1015                                 return err
1016                         }
1017                         if *mergeOutput {
1018                                 err = os.Remove(annotationsFilename)
1019                                 if err != nil {
1020                                         return err
1021                                 }
1022                         }
1023                         for _, line := range bytes.Split(buf, []byte{'\n'}) {
1024                                 if len(line) == 0 {
1025                                         continue
1026                                 }
1027                                 fields := bytes.SplitN(line, []byte{','}, 9)
1028                                 tag, _ := strconv.Atoi(string(fields[0]))
1029                                 incol, _ := strconv.Atoi(string(fields[1]))
1030                                 tileVariant, _ := strconv.Atoi(string(fields[2]))
1031                                 hgvsID := string(fields[3])
1032                                 seqname := string(fields[4])
1033                                 pos, _ := strconv.Atoi(string(fields[5]))
1034                                 refseq := fields[6]
1035                                 if hgvsID == "" {
1036                                         // Null entry for un-diffable
1037                                         // tile variant
1038                                         continue
1039                                 }
1040                                 if hgvsID == "=" {
1041                                         // Null entry for ref tile
1042                                         continue
1043                                 }
1044                                 if mask != nil && !mask.Check(strings.TrimPrefix(seqname, "chr"), pos, pos+len(refseq)) {
1045                                         // The tile intersects one of
1046                                         // the selected regions, but
1047                                         // this particular HGVS
1048                                         // variant does not.
1049                                         continue
1050                                 }
1051                                 hgvsColPair := hgvsCols[hgvsID]
1052                                 if hgvsColPair[0] == nil {
1053                                         // values in new columns start
1054                                         // out as -1 ("no data yet")
1055                                         // or 0 ("=ref") here, may
1056                                         // change to 1 ("hgvs variant
1057                                         // present") below, either on
1058                                         // this line or a future line.
1059                                         hgvsColPair = [2][]int16{make([]int16, len(cmd.cgnames)), make([]int16, len(cmd.cgnames))}
1060                                         rt, ok := reftile[tagID(tag)]
1061                                         if !ok {
1062                                                 err = fmt.Errorf("bug: seeing annotations for tag %d, but it has no reftile entry", tag)
1063                                                 return err
1064                                         }
1065                                         for ph := 0; ph < 2; ph++ {
1066                                                 for row := 0; row < rows; row++ {
1067                                                         v := chunk[row*chunkcols+incol*2+ph]
1068                                                         if tileVariantID(v) == rt.variant {
1069                                                                 hgvsColPair[ph][row] = 0
1070                                                         } else {
1071                                                                 hgvsColPair[ph][row] = -1
1072                                                         }
1073                                                 }
1074                                         }
1075                                         hgvsCols[hgvsID] = hgvsColPair
1076                                         if annow != nil {
1077                                                 hgvsref := hgvs.Variant{
1078                                                         Position: pos,
1079                                                         Ref:      string(refseq),
1080                                                         New:      string(refseq),
1081                                                 }
1082                                                 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])
1083                                         }
1084                                 }
1085                                 if annow != nil {
1086                                         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])
1087                                 }
1088                                 for ph := 0; ph < 2; ph++ {
1089                                         for row := 0; row < rows; row++ {
1090                                                 v := chunk[row*chunkcols+incol*2+ph]
1091                                                 if int(v) == tileVariant {
1092                                                         hgvsColPair[ph][row] = 1
1093                                                 }
1094                                         }
1095                                 }
1096                         }
1097
1098                         startcol += chunkcols
1099                 }
1100                 if *mergeOutput {
1101                         err = annow.Flush()
1102                         if err != nil {
1103                                 return err
1104                         }
1105                         err = annof.Close()
1106                         if err != nil {
1107                                 return err
1108                         }
1109                         err = writeNumpyInt16(fmt.Sprintf("%s/matrix.npy", *outputDir), out, rows, cols)
1110                         if err != nil {
1111                                 return err
1112                         }
1113                 }
1114                 out = nil
1115
1116                 if *hgvsSingle {
1117                         cols = len(hgvsCols) * 2
1118                         log.Printf("building hgvs-based matrix: %d rows x %d cols", rows, cols)
1119                         out = make([]int16, rows*cols)
1120                         hgvsIDs := make([]string, 0, cols/2)
1121                         for hgvsID := range hgvsCols {
1122                                 hgvsIDs = append(hgvsIDs, hgvsID)
1123                         }
1124                         sort.Strings(hgvsIDs)
1125                         var hgvsLabels bytes.Buffer
1126                         for idx, hgvsID := range hgvsIDs {
1127                                 fmt.Fprintf(&hgvsLabels, "%d,%s\n", idx, hgvsID)
1128                                 for ph := 0; ph < 2; ph++ {
1129                                         hgvscol := hgvsCols[hgvsID][ph]
1130                                         for row, val := range hgvscol {
1131                                                 out[row*cols+idx*2+ph] = val
1132                                         }
1133                                 }
1134                         }
1135                         err = writeNumpyInt16(fmt.Sprintf("%s/hgvs.npy", *outputDir), out, rows, cols)
1136                         if err != nil {
1137                                 return err
1138                         }
1139
1140                         fnm := fmt.Sprintf("%s/hgvs.annotations.csv", *outputDir)
1141                         log.Printf("writing hgvs labels: %s", fnm)
1142                         err = ioutil.WriteFile(fnm, hgvsLabels.Bytes(), 0777)
1143                         if err != nil {
1144                                 return err
1145                         }
1146                 }
1147         }
1148         if *onehotSingle || *onlyPCA {
1149                 nzCount := 0
1150                 for _, part := range onehotIndirect {
1151                         nzCount += len(part[0])
1152                 }
1153                 onehot := make([]uint32, nzCount*2) // [r,r,r,...,c,c,c,...]
1154                 var xrefs []onehotXref
1155                 chunkOffset := uint32(0)
1156                 outcol := 0
1157                 for i, part := range onehotIndirect {
1158                         for i := range part[1] {
1159                                 part[1][i] += chunkOffset
1160                         }
1161                         copy(onehot[outcol:], part[0])
1162                         copy(onehot[outcol+nzCount:], part[1])
1163                         xrefs = append(xrefs, onehotXrefs[i]...)
1164
1165                         outcol += len(part[0])
1166                         chunkOffset += onehotChunkSize[i]
1167
1168                         part[0] = nil
1169                         part[1] = nil
1170                         onehotXrefs[i] = nil
1171                         debug.FreeOSMemory()
1172                 }
1173                 if *onehotSingle {
1174                         fnm := fmt.Sprintf("%s/onehot.npy", *outputDir)
1175                         err = writeNumpyUint32(fnm, onehot, 2, nzCount)
1176                         if err != nil {
1177                                 return err
1178                         }
1179                         fnm = fmt.Sprintf("%s/onehot-columns.npy", *outputDir)
1180                         err = writeNumpyInt32(fnm, onehotXref2int32(xrefs), 5, len(xrefs))
1181                         if err != nil {
1182                                 return err
1183                         }
1184                 }
1185                 if *onlyPCA {
1186                         cols := 0
1187                         for _, c := range onehot[nzCount:] {
1188                                 if int(c) >= cols {
1189                                         cols = int(c) + 1
1190                                 }
1191                         }
1192                         if cols == 0 {
1193                                 return fmt.Errorf("cannot do PCA: one-hot matrix is empty")
1194                         }
1195                         log.Printf("have %d one-hot cols", cols)
1196                         stride := 1
1197                         for *maxPCATiles > 0 && cols > *maxPCATiles*2 {
1198                                 cols = (cols + 1) / 2
1199                                 stride = stride * 2
1200                         }
1201                         if cols%2 == 1 {
1202                                 // we work with pairs of columns
1203                                 cols++
1204                         }
1205                         log.Printf("creating full matrix (%d rows) and training matrix (%d rows) with %d cols, stride %d", len(cmd.cgnames), cmd.trainingSetSize, cols, stride)
1206                         mtxFull := mat.NewDense(len(cmd.cgnames), cols, nil)
1207                         mtxTrain := mat.NewDense(cmd.trainingSetSize, cols, nil)
1208                         for i, c := range onehot[nzCount:] {
1209                                 if int(c/2)%stride == 0 {
1210                                         outcol := int(c/2)/stride*2 + int(c)%2
1211                                         mtxFull.Set(int(onehot[i]), outcol, 1)
1212                                         if trainRow := cmd.trainingSet[int(onehot[i])]; trainRow >= 0 {
1213                                                 mtxTrain.Set(trainRow, outcol, 1)
1214                                         }
1215                                 }
1216                         }
1217                         log.Print("fitting")
1218                         transformer := nlp.NewPCA(cmd.pcaComponents)
1219                         transformer.Fit(mtxTrain.T())
1220                         log.Printf("transforming")
1221                         pca, err := transformer.Transform(mtxFull.T())
1222                         if err != nil {
1223                                 return err
1224                         }
1225                         pca = pca.T()
1226                         outrows, outcols := pca.Dims()
1227                         log.Printf("copying result to numpy output array: %d rows, %d cols", outrows, outcols)
1228                         out := make([]float64, outrows*outcols)
1229                         for i := 0; i < outrows; i++ {
1230                                 for j := 0; j < outcols; j++ {
1231                                         out[i*outcols+j] = pca.At(i, j)
1232                                 }
1233                         }
1234                         fnm := fmt.Sprintf("%s/pca.npy", *outputDir)
1235                         log.Printf("writing numpy: %s", fnm)
1236                         output, err := os.OpenFile(fnm, os.O_CREATE|os.O_TRUNC|os.O_WRONLY, 0777)
1237                         if err != nil {
1238                                 return err
1239                         }
1240                         npw, err := gonpy.NewWriter(nopCloser{output})
1241                         if err != nil {
1242                                 return fmt.Errorf("gonpy.NewWriter: %w", err)
1243                         }
1244                         npw.Shape = []int{outrows, outcols}
1245                         err = npw.WriteFloat64(out)
1246                         if err != nil {
1247                                 return fmt.Errorf("WriteFloat64: %w", err)
1248                         }
1249                         err = output.Close()
1250                         if err != nil {
1251                                 return err
1252                         }
1253                         log.Print("done")
1254
1255                         samplesOutFilename := *outputDir + "/samples.csv"
1256                         log.Infof("writing sample metadata to %s", samplesOutFilename)
1257                         var f *os.File
1258                         f, err = os.Create(samplesOutFilename)
1259                         if err != nil {
1260                                 return err
1261                         }
1262                         defer f.Close()
1263                         for i, si := range cmd.samples {
1264                                 var cc, tv string
1265                                 if si.isCase {
1266                                         cc = "1"
1267                                 } else if si.isControl {
1268                                         cc = "0"
1269                                 }
1270                                 if si.isTraining {
1271                                         tv = "1"
1272                                 } else {
1273                                         tv = "0"
1274                                 }
1275                                 var pcavals string
1276                                 for c := 0; c < outcols; c++ {
1277                                         pcavals += fmt.Sprintf(",%f", pca.At(i, c))
1278                                 }
1279                                 _, err = fmt.Fprintf(f, "%d,%s,%s,%s%s\n", i, si.id, cc, tv, pcavals)
1280                                 if err != nil {
1281                                         err = fmt.Errorf("write %s: %w", samplesOutFilename, err)
1282                                         return err
1283                                 }
1284                         }
1285                         err = f.Close()
1286                         if err != nil {
1287                                 err = fmt.Errorf("close %s: %w", samplesOutFilename, err)
1288                                 return err
1289                         }
1290                         log.Print("done")
1291                 }
1292         }
1293         if !*mergeOutput && !*onehotChunked && !*onehotSingle && !*onlyPCA {
1294                 tagoffsetFilename := *outputDir + "/chunk-tag-offset.csv"
1295                 log.Infof("writing tag offsets to %s", tagoffsetFilename)
1296                 var f *os.File
1297                 f, err = os.Create(tagoffsetFilename)
1298                 if err != nil {
1299                         return err
1300                 }
1301                 defer f.Close()
1302                 for idx, offset := range chunkStartTag {
1303                         _, err = fmt.Fprintf(f, "%q,%d\n", fmt.Sprintf("matrix.%04d.npy", idx), offset)
1304                         if err != nil {
1305                                 err = fmt.Errorf("write %s: %w", tagoffsetFilename, err)
1306                                 return err
1307                         }
1308                 }
1309                 err = f.Close()
1310                 if err != nil {
1311                         err = fmt.Errorf("close %s: %w", tagoffsetFilename, err)
1312                         return err
1313                 }
1314         }
1315
1316         return nil
1317 }
1318
1319 type sampleInfo struct {
1320         id            string
1321         isCase        bool
1322         isControl     bool
1323         isTraining    bool
1324         isValidation  bool
1325         pcaComponents []float64
1326 }
1327
1328 // Read samples.csv file with case/control and training/validation
1329 // flags.
1330 func loadSampleInfo(samplesFilename string) ([]sampleInfo, error) {
1331         var si []sampleInfo
1332         f, err := open(samplesFilename)
1333         if err != nil {
1334                 return nil, err
1335         }
1336         buf, err := io.ReadAll(f)
1337         f.Close()
1338         if err != nil {
1339                 return nil, err
1340         }
1341         lineNum := 0
1342         for _, csv := range bytes.Split(buf, []byte{'\n'}) {
1343                 lineNum++
1344                 if len(csv) == 0 {
1345                         continue
1346                 }
1347                 split := strings.Split(string(csv), ",")
1348                 if len(split) < 4 {
1349                         return nil, fmt.Errorf("%d fields < 4 in %s line %d: %q", len(split), samplesFilename, lineNum, csv)
1350                 }
1351                 if split[0] == "Index" && split[1] == "SampleID" && split[2] == "CaseControl" && split[3] == "TrainingValidation" {
1352                         continue
1353                 }
1354                 idx, err := strconv.Atoi(split[0])
1355                 if err != nil {
1356                         if lineNum == 1 {
1357                                 return nil, fmt.Errorf("header does not look right: %q", csv)
1358                         }
1359                         return nil, fmt.Errorf("%s line %d: index: %s", samplesFilename, lineNum, err)
1360                 }
1361                 if idx != len(si) {
1362                         return nil, fmt.Errorf("%s line %d: index %d out of order", samplesFilename, lineNum, idx)
1363                 }
1364                 var pcaComponents []float64
1365                 if len(split) > 4 {
1366                         for _, s := range split[4:] {
1367                                 f, err := strconv.ParseFloat(s, 64)
1368                                 if err != nil {
1369                                         return nil, fmt.Errorf("%s line %d: cannot parse float %q: %s", samplesFilename, lineNum, s, err)
1370                                 }
1371                                 pcaComponents = append(pcaComponents, f)
1372                         }
1373                 }
1374                 si = append(si, sampleInfo{
1375                         id:            split[1],
1376                         isCase:        split[2] == "1",
1377                         isControl:     split[2] == "0",
1378                         isTraining:    split[3] == "1",
1379                         isValidation:  split[3] == "0",
1380                         pcaComponents: pcaComponents,
1381                 })
1382         }
1383         return si, nil
1384 }
1385
1386 func (cmd *sliceNumpy) filterHGVScolpair(colpair [2][]int8) bool {
1387         if cmd.chi2PValue >= 1 {
1388                 return true
1389         }
1390         col0 := make([]bool, 0, len(cmd.chi2Cases))
1391         col1 := make([]bool, 0, len(cmd.chi2Cases))
1392         cases := make([]bool, 0, len(cmd.chi2Cases))
1393         for i, c := range cmd.chi2Cases {
1394                 if colpair[0][i] < 0 {
1395                         continue
1396                 }
1397                 col0 = append(col0, colpair[0][i] != 0)
1398                 col1 = append(col1, colpair[1][i] != 0)
1399                 cases = append(cases, c)
1400         }
1401         return len(cases) >= cmd.minCoverage &&
1402                 (pvalue(col0, cases) <= cmd.chi2PValue || pvalue(col1, cases) <= cmd.chi2PValue)
1403 }
1404
1405 func writeNumpyUint32(fnm string, out []uint32, 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.WriteUint32(out)
1424         err = bufw.Flush()
1425         if err != nil {
1426                 return err
1427         }
1428         return output.Close()
1429 }
1430
1431 func writeNumpyInt32(fnm string, out []int32, 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 * 4,
1447         }).Infof("writing numpy: %s", fnm)
1448         npw.Shape = []int{rows, cols}
1449         npw.WriteInt32(out)
1450         err = bufw.Flush()
1451         if err != nil {
1452                 return err
1453         }
1454         return output.Close()
1455 }
1456
1457 func writeNumpyInt16(fnm string, out []int16, 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 * 2,
1473         }).Infof("writing numpy: %s", fnm)
1474         npw.Shape = []int{rows, cols}
1475         npw.WriteInt16(out)
1476         err = bufw.Flush()
1477         if err != nil {
1478                 return err
1479         }
1480         return output.Close()
1481 }
1482
1483 func writeNumpyInt8(fnm string, out []int8, rows, cols int) error {
1484         output, err := os.Create(fnm)
1485         if err != nil {
1486                 return err
1487         }
1488         defer output.Close()
1489         bufw := bufio.NewWriterSize(output, 1<<26)
1490         npw, err := gonpy.NewWriter(nopCloser{bufw})
1491         if err != nil {
1492                 return err
1493         }
1494         log.WithFields(log.Fields{
1495                 "filename": fnm,
1496                 "rows":     rows,
1497                 "cols":     cols,
1498                 "bytes":    rows * cols,
1499         }).Infof("writing numpy: %s", fnm)
1500         npw.Shape = []int{rows, cols}
1501         npw.WriteInt8(out)
1502         err = bufw.Flush()
1503         if err != nil {
1504                 return err
1505         }
1506         return output.Close()
1507 }
1508
1509 func allele2homhet(colpair [2][]int8) {
1510         a, b := colpair[0], colpair[1]
1511         for i, av := range a {
1512                 bv := b[i]
1513                 if av < 0 || bv < 0 {
1514                         // no-call
1515                         a[i], b[i] = -1, -1
1516                 } else if av > 0 && bv > 0 {
1517                         // hom
1518                         a[i], b[i] = 1, 0
1519                 } else if av > 0 || bv > 0 {
1520                         // het
1521                         a[i], b[i] = 0, 1
1522                 } else {
1523                         // ref (or a different variant in same position)
1524                         // (this is a no-op) a[i], b[i] = 0, 0
1525                 }
1526         }
1527 }
1528
1529 type onehotXref struct {
1530         tag     tagID
1531         variant tileVariantID
1532         hom     bool
1533         pvalue  float64
1534 }
1535
1536 const onehotXrefSize = unsafe.Sizeof(onehotXref{})
1537
1538 // Build onehot matrix (m[tileVariantIndex][genome] == 0 or 1) for all
1539 // variants of a single tile/tag#.
1540 //
1541 // Return nil if no tile variant passes Χ² filter.
1542 func (cmd *sliceNumpy) tv2homhet(cgs map[string]CompactGenome, maxv tileVariantID, remap []tileVariantID, tag, chunkstarttag tagID, seq map[tagID][]TileVariant) ([][]int8, []onehotXref) {
1543         if tag == cmd.debugTag {
1544                 tv := make([]tileVariantID, len(cmd.cgnames)*2)
1545                 for i, name := range cmd.cgnames {
1546                         copy(tv[i*2:(i+1)*2], cgs[name].Variants[(tag-chunkstarttag)*2:])
1547                 }
1548                 log.WithFields(logrus.Fields{
1549                         "cgs[i].Variants[tag*2+j]": tv,
1550                         "maxv":                     maxv,
1551                         "remap":                    remap,
1552                         "tag":                      tag,
1553                         "chunkstarttag":            chunkstarttag,
1554                 }).Info("tv2homhet()")
1555         }
1556         if maxv < 1 || (maxv < 2 && !cmd.includeVariant1) {
1557                 // everyone has the most common variant (of the variants we don't drop)
1558                 return nil, nil
1559         }
1560         tagoffset := tag - chunkstarttag
1561         coverage := 0
1562         for _, cg := range cgs {
1563                 alleles := 0
1564                 for _, v := range cg.Variants[tagoffset*2 : tagoffset*2+2] {
1565                         if v > 0 && int(v) < len(seq[tag]) && len(seq[tag][v].Sequence) > 0 {
1566                                 alleles++
1567                         }
1568                 }
1569                 if alleles == 2 {
1570                         coverage++
1571                 }
1572         }
1573         if coverage < cmd.minCoverage {
1574                 return nil, nil
1575         }
1576         // "observed" array for p-value calculation (training set
1577         // only)
1578         obs := make([][]bool, (maxv+1)*2) // 2 slices (hom + het) for each variant#
1579         // one-hot output (all samples)
1580         outcols := make([][]int8, (maxv+1)*2)
1581         for i := range obs {
1582                 obs[i] = make([]bool, cmd.trainingSetSize)
1583                 outcols[i] = make([]int8, len(cmd.cgnames))
1584         }
1585         for cgid, name := range cmd.cgnames {
1586                 tsid := cmd.trainingSet[cgid]
1587                 cgvars := cgs[name].Variants[tagoffset*2:]
1588                 tv0, tv1 := remap[cgvars[0]], remap[cgvars[1]]
1589                 for v := tileVariantID(1); v <= maxv; v++ {
1590                         if tv0 == v && tv1 == v {
1591                                 if tsid >= 0 {
1592                                         obs[v*2][tsid] = true
1593                                 }
1594                                 outcols[v*2][cgid] = 1
1595                         } else if tv0 == v || tv1 == v {
1596                                 if tsid >= 0 {
1597                                         obs[v*2+1][tsid] = true
1598                                 }
1599                                 outcols[v*2+1][cgid] = 1
1600                         }
1601                 }
1602         }
1603         var onehot [][]int8
1604         var xref []onehotXref
1605         for col := 2; col < len(obs); col++ {
1606                 // col 0,1 correspond to tile variant 0, i.e.,
1607                 // no-call; col 2,3 correspond to the most common
1608                 // variant; so we (normally) start at col 4.
1609                 if col < 4 && !cmd.includeVariant1 {
1610                         continue
1611                 }
1612                 p := cmd.pvalue(obs[col])
1613                 if cmd.chi2PValue < 1 && !(p < cmd.chi2PValue) {
1614                         continue
1615                 }
1616                 onehot = append(onehot, outcols[col])
1617                 xref = append(xref, onehotXref{
1618                         tag:     tag,
1619                         variant: tileVariantID(col >> 1),
1620                         hom:     col&1 == 0,
1621                         pvalue:  p,
1622                 })
1623         }
1624         return onehot, xref
1625 }
1626
1627 // convert a []onehotXref with length N to a numpy-style []int32
1628 // matrix with N columns, one row per field of onehotXref struct.
1629 //
1630 // Hom/het row contains hom=0, het=1.
1631 //
1632 // P-value row contains 1000000x actual p-value.
1633 func onehotXref2int32(xrefs []onehotXref) []int32 {
1634         xcols := len(xrefs)
1635         xdata := make([]int32, 5*xcols)
1636         for i, xref := range xrefs {
1637                 xdata[i] = int32(xref.tag)
1638                 xdata[xcols+i] = int32(xref.variant)
1639                 if xref.hom {
1640                         xdata[xcols*2+i] = 1
1641                 }
1642                 xdata[xcols*3+i] = int32(xref.pvalue * 1000000)
1643                 xdata[xcols*4+i] = int32(-math.Log10(xref.pvalue) * 1000000)
1644         }
1645         return xdata
1646 }
1647
1648 // transpose onehot data from in[col][row] to numpy-style
1649 // out[row*cols+col].
1650 func onehotcols2int8(in [][]int8) []int8 {
1651         if len(in) == 0 {
1652                 return nil
1653         }
1654         cols := len(in)
1655         rows := len(in[0])
1656         out := make([]int8, rows*cols)
1657         for row := 0; row < rows; row++ {
1658                 outrow := out[row*cols:]
1659                 for col, incol := range in {
1660                         outrow[col] = incol[row]
1661                 }
1662         }
1663         return out
1664 }
1665
1666 // Return [2][]uint32{rowIndices, colIndices} indicating which
1667 // elements of matrixT[c][r] have non-zero values.
1668 func onehotChunk2Indirect(matrixT [][]int8) [2][]uint32 {
1669         var nz [2][]uint32
1670         for c, col := range matrixT {
1671                 for r, val := range col {
1672                         if val != 0 {
1673                                 nz[0] = append(nz[0], uint32(r))
1674                                 nz[1] = append(nz[1], uint32(c))
1675                         }
1676                 }
1677         }
1678         return nz
1679 }