export.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         "context"
  11         "errors"
  12         "flag"
  13         "fmt"
  14         "io"
  15         "net/http"
  16         _ "net/http/pprof"
  17         "os"
  18         "path/filepath"
  19         "runtime"
  20         "sort"
  21         "strconv"
  22         "strings"
  23         "sync"
  24         "time"
  25
  26         "git.arvados.org/arvados.git/sdk/go/arvados"
  27         "github.com/arvados/lightning/hgvs"
  28         "github.com/klauspost/pgzip"
  29         "github.com/kshedden/gonpy"
  30         "github.com/sirupsen/logrus"
  31         log "github.com/sirupsen/logrus"
  32 )
  33
  34 type tvVariant struct {
  35         hgvs.Variant
  36         librefs map[tileLibRef]bool
  37 }
  38
  39 type outputFormat interface {
  40         Filename() string
  41         PadLeft() bool
  42         Head(out io.Writer, cgs []CompactGenome, cases []bool, p float64) error
  43         Print(out io.Writer, seqname string, varslice []tvVariant) error
  44         Finish(outdir string, out io.Writer, seqname string) error
  45         MaxGoroutines() int
  46 }
  47
  48 var outputFormats = map[string]func() outputFormat{
  49         "hgvs-numpy": func() outputFormat {
  50                 return &formatHGVSNumpy{alleles: map[string][][]int8{}}
  51         },
  52         "hgvs-onehot": func() outputFormat { return formatHGVSOneHot{} },
  53         "hgvs":        func() outputFormat { return formatHGVS{} },
  54         "pvcf":        func() outputFormat { return formatPVCF{} },
  55         "vcf":         func() outputFormat { return formatVCF{} },
  56 }
  57
  58 type exporter struct {
  59         outputFormat   outputFormat
  60         outputPerChrom bool
  61         compress       bool
  62         maxTileSize    int
  63         filter         filter
  64         maxPValue      float64
  65         cases          []bool
  66 }
  67
  68 func (cmd *exporter) RunCommand(prog string, args []string, stdin io.Reader, stdout, stderr io.Writer) int {
  69         var err error
  70         defer func() {
  71                 if err != nil {
  72                         fmt.Fprintf(stderr, "%s\n", err)
  73                 }
  74         }()
  75         flags := flag.NewFlagSet("", flag.ContinueOnError)
  76         flags.SetOutput(stderr)
  77         pprof := flags.String("pprof", "", "serve Go profile data at http://`[addr]:port`")
  78         pprofdir := flags.String("pprof-dir", "", "write Go profile data to `directory` periodically")
  79         runlocal := flags.Bool("local", false, "run on local host (default: run in an arvados container)")
  80         projectUUID := flags.String("project", "", "project `UUID` for output data")
  81         priority := flags.Int("priority", 500, "container request priority")
  82         refname := flags.String("ref", "", "reference genome `name`")
  83         inputDir := flags.String("input-dir", ".", "input `directory`")
  84         cases := flags.String("cases", "", "file indicating which genomes are positive cases (for computing p-values)")
  85         flags.Float64Var(&cmd.maxPValue, "p-value", 1, "do chi square test and omit columns with p-value above this threshold")
  86         outputDir := flags.String("output-dir", ".", "output `directory`")
  87         outputFormatStr := flags.String("output-format", "hgvs", "output `format`: hgvs, pvcf, or vcf")
  88         outputBed := flags.String("output-bed", "", "also output bed `file`")
  89         flags.BoolVar(&cmd.outputPerChrom, "output-per-chromosome", true, "output one file per chromosome")
  90         flags.BoolVar(&cmd.compress, "z", false, "write gzip-compressed output files")
  91         labelsFilename := flags.String("output-labels", "", "also output genome labels csv `file`")
  92         flags.IntVar(&cmd.maxTileSize, "max-tile-size", 50000, "don't try to make annotations for tiles bigger than given `size`")
  93         cmd.filter.Flags(flags)
  94         err = flags.Parse(args)
  95         if err == flag.ErrHelp {
  96                 err = nil
  97                 return 0
  98         } else if err != nil {
  99                 return 2
 100         }
 101         if flag.NArg() > 0 {
 102                 err = fmt.Errorf("extra unparsed command line arguments: %q", flag.Args())
 103                 return 2
 104         }
 105
 106         if f, ok := outputFormats[*outputFormatStr]; !ok {
 107                 err = fmt.Errorf("invalid output format %q", *outputFormatStr)
 108                 return 2
 109         } else {
 110                 cmd.outputFormat = f()
 111         }
 112
 113         if *pprof != "" {
 114                 go func() {
 115                         log.Println(http.ListenAndServe(*pprof, nil))
 116                 }()
 117         }
 118         if *pprofdir != "" {
 119                 go writeProfilesPeriodically(*pprofdir)
 120         }
 121
 122         if !*runlocal {
 123                 if *outputDir != "." {
 124                         err = errors.New("cannot specify output directory in container mode: not implemented")
 125                         return 1
 126                 }
 127                 runner := arvadosContainerRunner{
 128                         Name:        "lightning export",
 129                         Client:      arvados.NewClientFromEnv(),
 130                         ProjectUUID: *projectUUID,
 131                         RAM:         750000000000,
 132                         VCPUs:       96,
 133                         Priority:    *priority,
 134                         APIAccess:   true,
 135                 }
 136                 err = runner.TranslatePaths(inputDir, cases)
 137                 if err != nil {
 138                         return 1
 139                 }
 140                 if *outputBed != "" {
 141                         if strings.Contains(*outputBed, "/") {
 142                                 err = fmt.Errorf("cannot use -output-bed filename %q containing '/' char", *outputBed)
 143                                 return 1
 144                         }
 145                         *outputBed = "/mnt/output/" + *outputBed
 146                 }
 147                 runner.Args = []string{"export", "-local=true",
 148                         "-pprof", ":6000",
 149                         "-pprof-dir", "/mnt/output",
 150                         "-ref", *refname,
 151                         "-cases", *cases,
 152                         "-p-value", fmt.Sprintf("%f", cmd.maxPValue),
 153                         "-output-format", *outputFormatStr,
 154                         "-output-bed", *outputBed,
 155                         "-output-labels", "/mnt/output/labels.csv",
 156                         "-output-per-chromosome=" + fmt.Sprintf("%v", cmd.outputPerChrom),
 157                         "-max-tile-size", fmt.Sprintf("%d", cmd.maxTileSize),
 158                         "-input-dir", *inputDir,
 159                         "-output-dir", "/mnt/output",
 160                         "-z=" + fmt.Sprintf("%v", cmd.compress),
 161                 }
 162                 runner.Args = append(runner.Args, cmd.filter.Args()...)
 163                 var output string
 164                 output, err = runner.Run()
 165                 if err != nil {
 166                         return 1
 167                 }
 168                 fmt.Fprintln(stdout, output)
 169                 return 0
 170         }
 171
 172         var cgs []CompactGenome
 173         tilelib := &tileLibrary{
 174                 retainNoCalls:       true,
 175                 retainTileSequences: true,
 176                 compactGenomes:      map[string][]tileVariantID{},
 177         }
 178         err = tilelib.LoadDir(context.Background(), *inputDir)
 179         if err != nil {
 180                 return 1
 181         }
 182
 183         refseq, ok := tilelib.refseqs[*refname]
 184         if !ok {
 185                 err = fmt.Errorf("reference name %q not found in input; have %v", *refname, func() (names []string) {
 186                         for name := range tilelib.refseqs {
 187                                 names = append(names, name)
 188                         }
 189                         return
 190                 }())
 191                 return 1
 192         }
 193
 194         log.Infof("filtering: %+v", cmd.filter)
 195         cmd.filter.Apply(tilelib)
 196
 197         names := cgnames(tilelib)
 198         for _, name := range names {
 199                 cgs = append(cgs, CompactGenome{Name: name, Variants: tilelib.compactGenomes[name]})
 200         }
 201         if *labelsFilename != "" {
 202                 log.Infof("writing labels to %s", *labelsFilename)
 203                 var f *os.File
 204                 f, err = os.OpenFile(*labelsFilename, os.O_CREATE|os.O_WRONLY, 0777)
 205                 if err != nil {
 206                         return 1
 207                 }
 208                 defer f.Close()
 209                 for i, name := range names {
 210                         _, err = fmt.Fprintf(f, "%d,%q,%q\n", i, trimFilenameForLabel(name), cmd.outputFormat.Filename())
 211                         if err != nil {
 212                                 err = fmt.Errorf("write %s: %w", *labelsFilename, err)
 213                                 return 1
 214                         }
 215                 }
 216                 err = f.Close()
 217                 if err != nil {
 218                         err = fmt.Errorf("close %s: %w", *labelsFilename, err)
 219                         return 1
 220                 }
 221         }
 222
 223         cmd.cases = make([]bool, len(names))
 224         if *cases != "" {
 225                 log.Infof("reading cases file: %s", *cases)
 226                 var f io.ReadCloser
 227                 f, err = open(*cases)
 228                 if err != nil {
 229                         return 1
 230                 }
 231                 defer f.Close()
 232                 var buf []byte
 233                 buf, err = io.ReadAll(f)
 234                 if err != nil {
 235                         return 1
 236                 }
 237                 for _, pattern := range bytes.Split(buf, []byte("\n")) {
 238                         if len(pattern) == 0 {
 239                                 continue
 240                         }
 241                         pattern := string(pattern)
 242                         idx := -1
 243                         for i, name := range names {
 244                                 if !strings.Contains(name, pattern) {
 245                                         continue
 246                                 } else if idx >= 0 {
 247                                         err = fmt.Errorf("pattern %q in cases file matches multiple genome IDs: %q, %q", pattern, names[idx], name)
 248                                         return 1
 249                                 } else {
 250                                         idx = i
 251                                 }
 252                         }
 253                         if idx < 0 {
 254                                 log.Warnf("pattern %q in cases file does not match any genome IDs", pattern)
 255                                 continue
 256                         }
 257                         cmd.cases[idx] = true
 258                 }
 259         }
 260
 261         var bedout io.Writer
 262         var bedfile *os.File
 263         var bedbufw *bufio.Writer
 264         if *outputBed != "" {
 265                 bedfile, err = os.OpenFile(*outputBed, os.O_CREATE|os.O_WRONLY, 0666)
 266                 if err != nil {
 267                         return 1
 268                 }
 269                 defer bedfile.Close()
 270                 bedbufw = bufio.NewWriterSize(bedfile, 16*1024*1024)
 271                 bedout = bedbufw
 272         }
 273
 274         err = cmd.export(*outputDir, bedout, tilelib, refseq, cgs)
 275         if err != nil {
 276                 return 1
 277         }
 278         if bedout != nil {
 279                 err = bedbufw.Flush()
 280                 if err != nil {
 281                         return 1
 282                 }
 283                 err = bedfile.Close()
 284                 if err != nil {
 285                         return 1
 286                 }
 287         }
 288         return 0
 289 }
 290
 291 func (cmd *exporter) export(outdir string, bedout io.Writer, tilelib *tileLibrary, refseq map[string][]tileLibRef, cgs []CompactGenome) error {
 292         var seqnames []string
 293         var missing []tileLibRef
 294         for seqname, librefs := range refseq {
 295                 seqnames = append(seqnames, seqname)
 296                 for _, libref := range librefs {
 297                         if libref.Variant != 0 && tilelib.TileVariantSequence(libref) == nil {
 298                                 missing = append(missing, libref)
 299                         }
 300                 }
 301         }
 302         sort.Strings(seqnames)
 303
 304         if len(missing) > 0 {
 305                 if limit := 100; len(missing) > limit {
 306                         log.Warnf("first %d missing tiles: %v", limit, missing[:limit])
 307                 } else {
 308                         log.Warnf("missing tiles: %v", missing)
 309                 }
 310                 return fmt.Errorf("%d needed tiles are missing from library", len(missing))
 311         }
 312
 313         outw := make([]io.WriteCloser, len(seqnames))
 314         bedw := make([]io.WriteCloser, len(seqnames))
 315
 316         var merges sync.WaitGroup
 317         merge := func(dst io.Writer, src []io.WriteCloser, label string) {
 318                 var mtx sync.Mutex
 319                 for i, seqname := range seqnames {
 320                         pr, pw := io.Pipe()
 321                         src[i] = pw
 322                         merges.Add(1)
 323                         seqname := seqname
 324                         go func() {
 325                                 defer merges.Done()
 326                                 log.Infof("writing %s %s", seqname, label)
 327                                 scanner := bufio.NewScanner(pr)
 328                                 for scanner.Scan() {
 329                                         mtx.Lock()
 330                                         dst.Write(scanner.Bytes())
 331                                         dst.Write([]byte{'\n'})
 332                                         mtx.Unlock()
 333                                 }
 334                                 log.Infof("writing %s %s done", seqname, label)
 335                         }()
 336                 }
 337         }
 338         if cmd.outputPerChrom {
 339                 for i, seqname := range seqnames {
 340                         fnm := filepath.Join(outdir, strings.Replace(cmd.outputFormat.Filename(), ".", "."+seqname+".", 1))
 341                         if cmd.compress {
 342                                 fnm += ".gz"
 343                         }
 344                         f, err := os.OpenFile(fnm, os.O_CREATE|os.O_WRONLY|os.O_TRUNC, 0666)
 345                         if err != nil {
 346                                 return err
 347                         }
 348                         defer f.Close()
 349                         log.Infof("writing %q", f.Name())
 350                         outw[i] = f
 351                         if cmd.compress {
 352                                 z := pgzip.NewWriter(f)
 353                                 defer z.Close()
 354                                 outw[i] = z
 355                         }
 356                         err = cmd.outputFormat.Head(outw[i], cgs, cmd.cases, cmd.maxPValue)
 357                         if err != nil {
 358                                 return err
 359                         }
 360                 }
 361         } else {
 362                 fnm := filepath.Join(outdir, cmd.outputFormat.Filename())
 363                 if cmd.compress {
 364                         fnm += ".gz"
 365                 }
 366                 f, err := os.OpenFile(fnm, os.O_CREATE|os.O_WRONLY|os.O_TRUNC, 0666)
 367                 if err != nil {
 368                         return err
 369                 }
 370                 defer f.Close()
 371                 log.Infof("writing %q", fnm)
 372                 var out io.Writer = f
 373                 if cmd.compress {
 374                         z := pgzip.NewWriter(out)
 375                         defer z.Close()
 376                         out = z
 377                 }
 378                 cmd.outputFormat.Head(out, cgs, cmd.cases, cmd.maxPValue)
 379                 merge(out, outw, "output")
 380         }
 381         if bedout != nil {
 382                 merge(bedout, bedw, "bed")
 383         }
 384
 385         throttle := throttle{Max: runtime.NumCPU()}
 386         if max := cmd.outputFormat.MaxGoroutines(); max > 0 {
 387                 throttle.Max = max
 388         }
 389         log.Infof("assembling %d sequences in %d goroutines", len(seqnames), throttle.Max)
 390         for seqidx, seqname := range seqnames {
 391                 seqidx, seqname := seqidx, seqname
 392                 outw := outw[seqidx]
 393                 bedw := bedw[seqidx]
 394                 throttle.Acquire()
 395                 go func() {
 396                         defer throttle.Release()
 397                         if bedw != nil {
 398                                 defer bedw.Close()
 399                         }
 400                         outwb := bufio.NewWriterSize(outw, 8*1024*1024)
 401                         eachVariant(bedw, tilelib.taglib.keylen, seqname, refseq[seqname], tilelib, cgs, cmd.outputFormat.PadLeft(), cmd.maxTileSize, func(varslice []tvVariant) {
 402                                 err := cmd.outputFormat.Print(outwb, seqname, varslice)
 403                                 throttle.Report(err)
 404                         })
 405                         err := cmd.outputFormat.Finish(outdir, outwb, seqname)
 406                         throttle.Report(err)
 407                         err = outwb.Flush()
 408                         throttle.Report(err)
 409                         err = outw.Close()
 410                         throttle.Report(err)
 411                 }()
 412         }
 413
 414         merges.Wait()
 415         throttle.Wait()
 416         return throttle.Err()
 417 }
 418
 419 // Align genome tiles to reference tiles, call callback func on each
 420 // variant, and (if bedw is not nil) write tile coverage to bedw.
 421 func eachVariant(bedw io.Writer, taglen int, seqname string, reftiles []tileLibRef, tilelib *tileLibrary, cgs []CompactGenome, padLeft bool, maxTileSize int, callback func(varslice []tvVariant)) {
 422         t0 := time.Now()
 423         progressbar := time.NewTicker(time.Minute)
 424         defer progressbar.Stop()
 425         var outmtx sync.Mutex
 426         defer outmtx.Lock()
 427         refpos := 0
 428         variantAt := map[int][]tvVariant{} // variantAt[chromOffset][genomeIndex*2+phase]
 429         for refstep, libref := range reftiles {
 430                 select {
 431                 case <-progressbar.C:
 432                         var eta interface{}
 433                         if refstep > 0 {
 434                                 fin := t0.Add(time.Duration(float64(time.Now().Sub(t0)) * float64(len(reftiles)) / float64(refstep)))
 435                                 eta = fmt.Sprintf("%v (%v)", fin.Format(time.RFC3339), fin.Sub(time.Now()))
 436                         } else {
 437                                 eta = "N/A"
 438                         }
 439                         log.Printf("exportSeq: %s: refstep %d of %d, %.0f/s, ETA %v", seqname, refstep, len(reftiles), float64(refstep)/time.Now().Sub(t0).Seconds(), eta)
 440                 default:
 441                 }
 442                 diffs := map[tileLibRef][]hgvs.Variant{}
 443                 refseq := tilelib.TileVariantSequence(libref)
 444                 tagcoverage := 0 // number of times the start tag was found in genomes -- max is len(cgs)*2
 445                 for cgidx, cg := range cgs {
 446                         for phase := 0; phase < 2; phase++ {
 447                                 var variant tileVariantID
 448                                 if i := int(libref.Tag)*2 + phase; len(cg.Variants) > i {
 449                                         variant = cg.Variants[i]
 450                                 }
 451                                 if variant > 0 {
 452                                         tagcoverage++
 453                                 }
 454                                 if variant == libref.Variant || variant == 0 {
 455                                         continue
 456                                 }
 457                                 glibref := tileLibRef{Tag: libref.Tag, Variant: variant}
 458                                 vars, ok := diffs[glibref]
 459                                 if !ok {
 460                                         genomeseq := tilelib.TileVariantSequence(glibref)
 461                                         if len(genomeseq) == 0 {
 462                                                 // Hash is known but sequence
 463                                                 // is not, e.g., retainNoCalls
 464                                                 // was false during import
 465                                                 continue
 466                                         }
 467                                         if len(genomeseq) > maxTileSize {
 468                                                 continue
 469                                         }
 470                                         refSequence := refseq
 471                                         // If needed, extend the
 472                                         // reference sequence up to
 473                                         // the tag at the end of the
 474                                         // genomeseq sequence.
 475                                         refstepend := refstep + 1
 476                                         for refstepend < len(reftiles) && len(refSequence) >= taglen && !bytes.EqualFold(refSequence[len(refSequence)-taglen:], genomeseq[len(genomeseq)-taglen:]) && len(refSequence) <= maxTileSize {
 477                                                 if &refSequence[0] == &refseq[0] {
 478                                                         refSequence = append([]byte(nil), refSequence...)
 479                                                 }
 480                                                 refSequence = append(refSequence, tilelib.TileVariantSequence(reftiles[refstepend])...)
 481                                                 refstepend++
 482                                         }
 483                                         // (TODO: handle no-calls)
 484                                         if len(refSequence) <= maxTileSize {
 485                                                 refstr := strings.ToUpper(string(refSequence))
 486                                                 genomestr := strings.ToUpper(string(genomeseq))
 487                                                 vars, _ = hgvs.Diff(refstr, genomestr, time.Second)
 488                                         }
 489                                         diffs[glibref] = vars
 490                                 }
 491                                 for _, v := range vars {
 492                                         if padLeft {
 493                                                 v = v.PadLeft()
 494                                         }
 495                                         v.Position += refpos
 496                                         varslice := variantAt[v.Position]
 497                                         if varslice == nil {
 498                                                 varslice = make([]tvVariant, len(cgs)*2)
 499                                                 variantAt[v.Position] = varslice
 500                                         }
 501                                         varslice[cgidx*2+phase].Variant = v
 502                                         if varslice[cgidx*2+phase].librefs == nil {
 503                                                 varslice[cgidx*2+phase].librefs = map[tileLibRef]bool{glibref: true}
 504                                         } else {
 505                                                 varslice[cgidx*2+phase].librefs[glibref] = true
 506                                         }
 507                                 }
 508                         }
 509                 }
 510                 refpos += len(refseq) - taglen
 511
 512                 // Flush entries from variantAt that are behind
 513                 // refpos. Flush all entries if this is the last
 514                 // reftile of the path/chromosome.
 515                 flushpos := make([]int, 0, len(variantAt))
 516                 lastrefstep := refstep == len(reftiles)-1
 517                 for pos := range variantAt {
 518                         if lastrefstep || pos <= refpos {
 519                                 flushpos = append(flushpos, pos)
 520                         }
 521                 }
 522                 sort.Slice(flushpos, func(i, j int) bool { return flushpos[i] < flushpos[j] })
 523                 flushvariants := make([][]tvVariant, len(flushpos))
 524                 for i, pos := range flushpos {
 525                         varslice := variantAt[pos]
 526                         delete(variantAt, pos)
 527                         // Check for uninitialized (zero-value)
 528                         // elements in varslice
 529                         for i := range varslice {
 530                                 if varslice[i].Position != 0 {
 531                                         // Not a zero-value element
 532                                         continue
 533                                 }
 534                                 // Set the position so
 535                                 // varslice[*].Position are all equal
 536                                 varslice[i].Position = pos
 537                                 // This could be either =ref or a
 538                                 // missing/low-quality tile. Figure
 539                                 // out which.
 540                                 vidx := int(libref.Tag)*2 + i%2
 541                                 if vidx >= len(cgs[i/2].Variants) {
 542                                         // Missing tile.
 543                                         varslice[i].New = "-"
 544                                         continue
 545                                 }
 546                                 v := cgs[i/2].Variants[vidx]
 547                                 if v < 1 || len(tilelib.TileVariantSequence(tileLibRef{Tag: libref.Tag, Variant: v})) == 0 {
 548                                         // Missing/low-quality tile.
 549                                         varslice[i].New = "-" // fasta "gap of indeterminate length"
 550                                 }
 551                         }
 552                         flushvariants[i] = varslice
 553                 }
 554                 outmtx.Lock()
 555                 go func() {
 556                         defer outmtx.Unlock()
 557                         for _, varslice := range flushvariants {
 558                                 callback(varslice)
 559                         }
 560                 }()
 561                 if bedw != nil && len(refseq) > 0 {
 562                         tilestart := refpos - len(refseq) + taglen
 563                         tileend := refpos
 564                         if !lastrefstep {
 565                                 tileend += taglen
 566                         }
 567                         thickstart := tilestart + taglen
 568                         if refstep == 0 {
 569                                 thickstart = 0
 570                         }
 571                         thickend := refpos
 572
 573                         // coverage score, 0 to 1000
 574                         score := 1000
 575                         if len(cgs) > 0 {
 576                                 score = 1000 * tagcoverage / len(cgs) / 2
 577                         }
 578
 579                         fmt.Fprintf(bedw, "%s %d %d %d %d . %d %d\n",
 580                                 seqname, tilestart, tileend,
 581                                 libref.Tag,
 582                                 score,
 583                                 thickstart, thickend)
 584                 }
 585         }
 586 }
 587
 588 func bucketVarsliceByRef(varslice []tvVariant) map[string]map[string]int {
 589         byref := map[string]map[string]int{}
 590         for _, v := range varslice {
 591                 if v.Ref == "" && v.New == "" {
 592                         // =ref
 593                         continue
 594                 }
 595                 if v.New == "-" {
 596                         // no-call
 597                         continue
 598                 }
 599                 alts := byref[v.Ref]
 600                 if alts == nil {
 601                         alts = map[string]int{}
 602                         byref[v.Ref] = alts
 603                 }
 604                 alts[v.New]++
 605         }
 606         return byref
 607 }
 608
 609 type formatVCF struct{}
 610
 611 func (formatVCF) MaxGoroutines() int                     { return 0 }
 612 func (formatVCF) Filename() string                       { return "out.vcf" }
 613 func (formatVCF) PadLeft() bool                          { return true }
 614 func (formatVCF) Finish(string, io.Writer, string) error { return nil }
 615 func (formatVCF) Head(out io.Writer, cgs []CompactGenome, cases []bool, p float64) error {
 616         _, err := fmt.Fprint(out, "#CHROM\tPOS\tID\tREF\tALT\tQUAL\tFILTER\tINFO\n")
 617         return err
 618 }
 619 func (formatVCF) Print(out io.Writer, seqname string, varslice []tvVariant) error {
 620         for ref, alts := range bucketVarsliceByRef(varslice) {
 621                 altslice := make([]string, 0, len(alts))
 622                 for alt := range alts {
 623                         altslice = append(altslice, alt)
 624                 }
 625                 sort.Strings(altslice)
 626
 627                 info := "AC="
 628                 for i, a := range altslice {
 629                         if i > 0 {
 630                                 info += ","
 631                         }
 632                         info += strconv.Itoa(alts[a])
 633                 }
 634                 _, err := fmt.Fprintf(out, "%s\t%d\t.\t%s\t%s\t.\t.\t%s\n", seqname, varslice[0].Position, ref, strings.Join(altslice, ","), info)
 635                 if err != nil {
 636                         return err
 637                 }
 638         }
 639         return nil
 640 }
 641
 642 type formatPVCF struct{}
 643
 644 func (formatPVCF) MaxGoroutines() int                     { return 0 }
 645 func (formatPVCF) Filename() string                       { return "out.vcf" }
 646 func (formatPVCF) PadLeft() bool                          { return true }
 647 func (formatPVCF) Finish(string, io.Writer, string) error { return nil }
 648 func (formatPVCF) Head(out io.Writer, cgs []CompactGenome, cases []bool, p float64) error {
 649         fmt.Fprintln(out, `##FORMAT=<ID=GT,Number=1,Type=String,Description="Genotype">`)
 650         fmt.Fprintf(out, "#CHROM\tPOS\tID\tREF\tALT\tQUAL\tFILTER\tINFO\tFORMAT")
 651         for _, cg := range cgs {
 652                 fmt.Fprintf(out, "\t%s", cg.Name)
 653         }
 654         _, err := fmt.Fprintf(out, "\n")
 655         return err
 656 }
 657
 658 func (formatPVCF) Print(out io.Writer, seqname string, varslice []tvVariant) error {
 659         for ref, alts := range bucketVarsliceByRef(varslice) {
 660                 altslice := make([]string, 0, len(alts))
 661                 for alt := range alts {
 662                         altslice = append(altslice, alt)
 663                 }
 664                 sort.Strings(altslice)
 665                 for i, a := range altslice {
 666                         alts[a] = i + 1
 667                 }
 668                 _, err := fmt.Fprintf(out, "%s\t%d\t.\t%s\t%s\t.\t.\t.\tGT", seqname, varslice[0].Position, ref, strings.Join(altslice, ","))
 669                 if err != nil {
 670                         return err
 671                 }
 672                 for i := 0; i < len(varslice); i += 2 {
 673                         v1, v2 := varslice[i], varslice[i+1]
 674                         a1, a2 := alts[v1.New], alts[v2.New]
 675                         if v1.Ref != ref {
 676                                 // variant on allele 0 belongs on a
 677                                 // different output line -- same
 678                                 // chr,pos but different "ref" length
 679                                 a1 = 0
 680                         }
 681                         if v2.Ref != ref {
 682                                 a2 = 0
 683                         }
 684                         _, err := fmt.Fprintf(out, "\t%d/%d", a1, a2)
 685                         if err != nil {
 686                                 return err
 687                         }
 688                 }
 689                 _, err = out.Write([]byte{'\n'})
 690                 if err != nil {
 691                         return err
 692                 }
 693         }
 694         return nil
 695 }
 696
 697 type formatHGVS struct{}
 698
 699 func (formatHGVS) MaxGoroutines() int                                                     { return 0 }
 700 func (formatHGVS) Filename() string                                                       { return "out.tsv" }
 701 func (formatHGVS) PadLeft() bool                                                          { return false }
 702 func (formatHGVS) Head(out io.Writer, cgs []CompactGenome, cases []bool, p float64) error { return nil }
 703 func (formatHGVS) Finish(string, io.Writer, string) error                                 { return nil }
 704 func (formatHGVS) Print(out io.Writer, seqname string, varslice []tvVariant) error {
 705         for i := 0; i < len(varslice)/2; i++ {
 706                 if i > 0 {
 707                         out.Write([]byte{'\t'})
 708                 }
 709                 var1, var2 := varslice[i*2], varslice[i*2+1]
 710                 if var1.New == "-" || var2.New == "-" {
 711                         _, err := out.Write([]byte{'N'})
 712                         if err != nil {
 713                                 return err
 714                         }
 715                         continue
 716                 }
 717                 if var1.Variant == var2.Variant {
 718                         if var1.Ref == var1.New {
 719                                 _, err := out.Write([]byte{'.'})
 720                                 if err != nil {
 721                                         return err
 722                                 }
 723                         } else {
 724                                 _, err := fmt.Fprintf(out, "%s:g.%s", seqname, var1.String())
 725                                 if err != nil {
 726                                         return err
 727                                 }
 728                         }
 729                 } else {
 730                         _, err := fmt.Fprintf(out, "%s:g.[%s];[%s]", seqname, var1.String(), var2.String())
 731                         if err != nil {
 732                                 return err
 733                         }
 734                 }
 735         }
 736         _, err := out.Write([]byte{'\n'})
 737         return err
 738 }
 739
 740 type formatHGVSOneHot struct{}
 741
 742 func (formatHGVSOneHot) MaxGoroutines() int { return 0 }
 743 func (formatHGVSOneHot) Filename() string   { return "out.tsv" }
 744 func (formatHGVSOneHot) PadLeft() bool      { return false }
 745 func (formatHGVSOneHot) Head(out io.Writer, cgs []CompactGenome, cases []bool, p float64) error {
 746         return nil
 747 }
 748 func (formatHGVSOneHot) Finish(string, io.Writer, string) error { return nil }
 749 func (formatHGVSOneHot) Print(out io.Writer, seqname string, varslice []tvVariant) error {
 750         vars := map[hgvs.Variant]bool{}
 751         for _, v := range varslice {
 752                 if v.Ref != v.New {
 753                         vars[v.Variant] = true
 754                 }
 755         }
 756
 757         // sort variants to ensure output is deterministic
 758         sorted := make([]hgvs.Variant, 0, len(vars))
 759         for v := range vars {
 760                 sorted = append(sorted, v)
 761         }
 762         sort.Slice(sorted, func(a, b int) bool { return hgvs.Less(sorted[a], sorted[b]) })
 763
 764         for _, v := range sorted {
 765                 if v.New == "-" {
 766                         continue
 767                 }
 768                 fmt.Fprintf(out, "%s.%s", seqname, v.String())
 769                 for i := 0; i < len(varslice); i += 2 {
 770                         if varslice[i].Variant == v || varslice[i+1].Variant == v {
 771                                 out.Write([]byte("\t1"))
 772                         } else {
 773                                 out.Write([]byte("\t0"))
 774                         }
 775                 }
 776                 _, err := out.Write([]byte{'\n'})
 777                 if err != nil {
 778                         return err
 779                 }
 780         }
 781         return nil
 782 }
 783
 784 type formatHGVSNumpy struct {
 785         sync.Mutex
 786         writelock sync.Mutex
 787         alleles   map[string][][]int8 // alleles[seqname][variantidx][genomeidx*2+phase]
 788         cases     []bool
 789         maxPValue float64
 790 }
 791
 792 func (*formatHGVSNumpy) MaxGoroutines() int { return 4 }
 793 func (*formatHGVSNumpy) Filename() string   { return "annotations.csv" }
 794 func (*formatHGVSNumpy) PadLeft() bool      { return false }
 795 func (f *formatHGVSNumpy) Head(out io.Writer, cgs []CompactGenome, cases []bool, p float64) error {
 796         f.cases = cases
 797         f.maxPValue = p
 798         return nil
 799 }
 800 func (f *formatHGVSNumpy) Print(outw io.Writer, seqname string, varslice []tvVariant) error {
 801         // sort variants to ensure output is deterministic
 802         sorted := make([]hgvs.Variant, 0, len(varslice))
 803         for _, v := range varslice {
 804                 sorted = append(sorted, v.Variant)
 805         }
 806         sort.Slice(sorted, func(a, b int) bool { return hgvs.Less(sorted[a], sorted[b]) })
 807
 808         f.Lock()
 809         seqalleles := f.alleles[seqname]
 810         f.Unlock()
 811
 812         chi2x := make([]bool, 0, len(varslice))
 813         chi2y := make([]bool, 0, len(varslice))
 814
 815         // append a row to seqalleles for each unique non-ref variant
 816         // in varslice.
 817         var previous hgvs.Variant
 818         for _, v := range sorted {
 819                 if previous == v || v.Ref == v.New || v.New == "-" {
 820                         continue
 821                 }
 822                 previous = v
 823                 chi2x, chi2y := chi2x, chi2y
 824                 newrow := make([]int8, len(varslice))
 825                 for i, allele := range varslice {
 826                         if allele.Variant == v {
 827                                 newrow[i] = 1
 828                                 chi2x = append(chi2x, true)
 829                                 chi2y = append(chi2y, f.cases[i/2])
 830                         } else if allele.Variant.New == "-" {
 831                                 newrow[i] = -1
 832                         } else {
 833                                 chi2x = append(chi2x, false)
 834                                 chi2y = append(chi2y, f.cases[i/2])
 835                         }
 836                 }
 837                 if f.maxPValue < 1 && pvalue(chi2x, chi2y) > f.maxPValue {
 838                         continue
 839                 }
 840                 seqalleles = append(seqalleles, newrow)
 841                 _, err := fmt.Fprintf(outw, "%d,%q\n", len(seqalleles)-1, seqname+"."+v.String())
 842                 if err != nil {
 843                         return err
 844                 }
 845         }
 846
 847         f.Lock()
 848         f.alleles[seqname] = seqalleles
 849         f.Unlock()
 850         return nil
 851 }
 852 func (f *formatHGVSNumpy) Finish(outdir string, _ io.Writer, seqname string) error {
 853         // Write seqname's data to a .npy matrix with one row per
 854         // genome and 2 columns per variant.
 855         f.Lock()
 856         seqalleles := f.alleles[seqname]
 857         delete(f.alleles, seqname)
 858         f.Unlock()
 859         if len(seqalleles) == 0 {
 860                 return nil
 861         }
 862         out := make([]int8, len(seqalleles)*len(seqalleles[0]))
 863         rows := len(seqalleles[0]) / 2
 864         cols := len(seqalleles) * 2
 865         // copy seqalleles[varidx][genome*2+phase] to
 866         // out[genome*nvars*2 + varidx*2 + phase]
 867         for varidx, alleles := range seqalleles {
 868                 for g := 0; g < len(alleles)/2; g++ {
 869                         aa, ab := alleles[g*2], alleles[g*2+1]
 870                         if aa < 0 || ab < 0 {
 871                                 // no-call
 872                                 out[g*cols+varidx*2] = -1
 873                                 out[g*cols+varidx*2+1] = -1
 874                         } else if aa > 0 && ab > 0 {
 875                                 // hom
 876                                 out[g*cols+varidx*2] = 1
 877                         } else if aa > 0 || ab > 0 {
 878                                 // het
 879                                 out[g*cols+varidx*2+1] = 1
 880                         }
 881                 }
 882         }
 883         outf, err := os.OpenFile(outdir+"/matrix."+seqname+".npy", os.O_CREATE|os.O_EXCL|os.O_WRONLY, 0777)
 884         if err != nil {
 885                 return err
 886         }
 887         defer outf.Close()
 888         bufw := bufio.NewWriter(outf)
 889         npw, err := gonpy.NewWriter(nopCloser{bufw})
 890         if err != nil {
 891                 return err
 892         }
 893         log.WithFields(logrus.Fields{
 894                 "seqname": seqname,
 895                 "rows":    rows,
 896                 "cols":    cols,
 897         }).Info("writing numpy")
 898         npw.Shape = []int{rows, cols}
 899         f.writelock.Lock() // serialize because WriteInt8 uses lots of memory
 900         npw.WriteInt8(out)
 901         f.writelock.Unlock()
 902         err = bufw.Flush()
 903         if err != nil {
 904                 return err
 905         }
 906         err = outf.Close()
 907         if err != nil {
 908                 return err
 909         }
 910         return nil
 911 }