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