exportnumpy.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         "io/ioutil"
  16         "net/http"
  17         _ "net/http/pprof"
  18         "os"
  19         "sort"
  20         "strconv"
  21         "strings"
  22         "sync"
  23         "sync/atomic"
  24
  25         "git.arvados.org/arvados.git/sdk/go/arvados"
  26         "github.com/arvados/lightning/hgvs"
  27         "github.com/kshedden/gonpy"
  28         "github.com/sirupsen/logrus"
  29         log "github.com/sirupsen/logrus"
  30 )
  31
  32 type exportNumpy struct {
  33         filter filter
  34 }
  35
  36 func (cmd *exportNumpy) RunCommand(prog string, args []string, stdin io.Reader, stdout, stderr io.Writer) int {
  37         var err error
  38         defer func() {
  39                 if err != nil {
  40                         fmt.Fprintf(stderr, "%s\n", err)
  41                 }
  42         }()
  43         flags := flag.NewFlagSet("", flag.ContinueOnError)
  44         flags.SetOutput(stderr)
  45         pprof := flags.String("pprof", "", "serve Go profile data at http://`[addr]:port`")
  46         runlocal := flags.Bool("local", false, "run on local host (default: run in an arvados container)")
  47         projectUUID := flags.String("project", "", "project `UUID` for output data")
  48         priority := flags.Int("priority", 500, "container request priority")
  49         inputDir := flags.String("input-dir", "./in", "input `directory`")
  50         outputDir := flags.String("output-dir", "./out", "output `directory`")
  51         annotationsFilename := flags.String("output-annotations", "", "output `file` for tile variant annotations csv")
  52         librefsFilename := flags.String("output-onehot2tilevar", "", "when using -one-hot, create csv `file` mapping column# to tag# and variant#")
  53         labelsFilename := flags.String("output-labels", "", "output `file` for genome labels csv")
  54         regionsFilename := flags.String("regions", "", "only output columns/annotations that intersect regions in specified bed `file`")
  55         expandRegions := flags.Int("expand-regions", 0, "expand specified regions by `N` base pairs on each side`")
  56         onehot := flags.Bool("one-hot", false, "recode tile variants as one-hot")
  57         chunks := flags.Int("chunks", 1, "split output into `N` numpy files")
  58         cmd.filter.Flags(flags)
  59         err = flags.Parse(args)
  60         if err == flag.ErrHelp {
  61                 err = nil
  62                 return 0
  63         } else if err != nil {
  64                 return 2
  65         }
  66
  67         if *pprof != "" {
  68                 go func() {
  69                         log.Println(http.ListenAndServe(*pprof, nil))
  70                 }()
  71         }
  72
  73         if !*runlocal {
  74                 runner := arvadosContainerRunner{
  75                         Name:        "lightning export-numpy",
  76                         Client:      arvados.NewClientFromEnv(),
  77                         ProjectUUID: *projectUUID,
  78                         RAM:         500000000000,
  79                         VCPUs:       96,
  80                         Priority:    *priority,
  81                         KeepCache:   1,
  82                         APIAccess:   true,
  83                 }
  84                 err = runner.TranslatePaths(inputDir, regionsFilename)
  85                 if err != nil {
  86                         return 1
  87                 }
  88                 runner.Args = []string{"export-numpy", "-local=true",
  89                         "-pprof", ":6060",
  90                         fmt.Sprintf("-one-hot=%v", *onehot),
  91                         "-input-dir", *inputDir,
  92                         "-output-dir", "/mnt/output",
  93                         "-output-annotations", "/mnt/output/annotations.csv",
  94                         "-output-onehot2tilevar", "/mnt/output/onehot2tilevar.csv",
  95                         "-output-labels", "/mnt/output/labels.csv",
  96                         "-regions", *regionsFilename,
  97                         "-expand-regions", fmt.Sprintf("%d", *expandRegions),
  98                         "-chunks", fmt.Sprintf("%d", *chunks),
  99                 }
 100                 runner.Args = append(runner.Args, cmd.filter.Args()...)
 101                 var output string
 102                 output, err = runner.Run()
 103                 if err != nil {
 104                         return 1
 105                 }
 106                 fmt.Fprintln(stdout, output+"/matrix.npy")
 107                 return 0
 108         }
 109
 110         tilelib := &tileLibrary{
 111                 retainNoCalls:       true,
 112                 retainTileSequences: true,
 113                 compactGenomes:      map[string][]tileVariantID{},
 114         }
 115         err = tilelib.LoadDir(context.Background(), *inputDir)
 116         if err != nil {
 117                 return 1
 118         }
 119
 120         log.Info("filtering")
 121         cmd.filter.Apply(tilelib)
 122         log.Info("tidying")
 123         tilelib.Tidy()
 124
 125         log.Info("building lowqual map")
 126         lowqual := lowqual(tilelib)
 127         names := cgnames(tilelib)
 128
 129         if *labelsFilename != "" {
 130                 log.Infof("writing labels to %s", *labelsFilename)
 131                 var f *os.File
 132                 f, err = os.OpenFile(*labelsFilename, os.O_CREATE|os.O_WRONLY, 0777)
 133                 if err != nil {
 134                         return 1
 135                 }
 136                 defer f.Close()
 137                 outBasename := "matrix.npy"
 138                 for i, name := range names {
 139                         _, err = fmt.Fprintf(f, "%d,%q,%q\n", i, trimFilenameForLabel(name), outBasename)
 140                         if err != nil {
 141                                 err = fmt.Errorf("write %s: %w", *labelsFilename, err)
 142                                 return 1
 143                         }
 144                 }
 145                 err = f.Close()
 146                 if err != nil {
 147                         err = fmt.Errorf("close %s: %w", *labelsFilename, err)
 148                         return 1
 149                 }
 150         }
 151
 152         log.Info("determining which tiles intersect given regions")
 153         dropTiles, err := chooseTiles(tilelib, *regionsFilename, *expandRegions)
 154         if err != nil {
 155                 return 1
 156         }
 157
 158         annotation2tvs := map[string]map[hgvs.Variant][]tileLibRef{}
 159         if *annotationsFilename != "" {
 160                 log.Info("writing annotations")
 161                 var annow io.WriteCloser
 162                 annow, err = os.OpenFile(*annotationsFilename, os.O_CREATE|os.O_WRONLY, 0666)
 163                 if err != nil {
 164                         return 1
 165                 }
 166                 defer annow.Close()
 167                 var mtx sync.Mutex
 168                 err = (&annotatecmd{
 169                         maxTileSize: 5000,
 170                         dropTiles:   dropTiles,
 171                         reportAnnotation: func(tag tagID, _ int, variant tileVariantID, refname string, seqname string, pdi hgvs.Variant) {
 172                                 mtx.Lock()
 173                                 defer mtx.Unlock()
 174                                 if annotation2tvs[seqname] == nil {
 175                                         annotation2tvs[seqname] = map[hgvs.Variant][]tileLibRef{}
 176                                 }
 177                                 annotation2tvs[seqname][pdi] = append(annotation2tvs[seqname][pdi], tileLibRef{Tag: tag, Variant: variant})
 178                         },
 179                 }).exportTileDiffs(annow, tilelib)
 180                 if err != nil {
 181                         return 1
 182                 }
 183                 err = annow.Close()
 184                 if err != nil {
 185                         return 1
 186                 }
 187         }
 188
 189         var lastErr atomic.Value
 190         var wg sync.WaitGroup
 191         for seqname, pdivars := range annotation2tvs {
 192                 seqname, pdivars := seqname, pdivars
 193                 wg.Add(1)
 194                 go func() {
 195                         defer wg.Done()
 196                         log.Infof("choosing hgvs columns for seq %s", seqname)
 197                         var pdis []hgvs.Variant
 198                         for pdi, librefs := range pdivars {
 199                                 // Include this HGVS column if it was
 200                                 // seen in a variant of any
 201                                 // non-dropped tile.
 202                                 for _, libref := range librefs {
 203                                         if int(libref.Tag) >= len(dropTiles) || !dropTiles[libref.Tag] {
 204                                                 pdis = append(pdis, pdi)
 205                                                 break
 206                                         }
 207                                 }
 208                         }
 209                         sort.Slice(pdis, func(i, j int) bool {
 210                                 if cmp := pdis[i].Position - pdis[j].Position; cmp != 0 {
 211                                         return cmp < 0
 212                                 } else if pdis[i].Ref != pdis[j].Ref {
 213                                         return pdis[i].Ref < pdis[j].Ref
 214                                 } else {
 215                                         return pdis[i].New < pdis[j].New
 216                                 }
 217                         })
 218                         log.Infof("writing column labels for seq %s", seqname)
 219                         var buf bytes.Buffer
 220                         for _, pdi := range pdis {
 221                                 fmt.Fprintf(&buf, "%s:g.%s\n", seqname, pdi.String())
 222                         }
 223                         err := ioutil.WriteFile(*outputDir+"/"+seqname+".columns.csv", buf.Bytes(), 0777)
 224                         if err != nil {
 225                                 lastErr.Store(err)
 226                                 return
 227                         }
 228
 229                         log.Infof("building hgvs matrix for seq %s", seqname)
 230                         data := make([]int8, len(names)*len(pdis)*2)
 231                         for row, name := range names {
 232                                 cg := tilelib.compactGenomes[name]
 233                                 rowstart := row * len(pdis) * 2
 234                                 for col, pdi := range pdis {
 235                                         for _, libref := range pdivars[pdi] {
 236                                                 if len(cg) <= int(libref.Tag)*2+1 {
 237                                                         continue
 238                                                 }
 239                                                 for phase := 0; phase < 2; phase++ {
 240                                                         if cg[int(libref.Tag)*2+phase] == libref.Variant {
 241                                                                 data[rowstart+col*2+phase] = 1
 242                                                         }
 243                                                 }
 244                                         }
 245                                 }
 246                         }
 247                         log.Infof("writing hgvs numpy for seq %s", seqname)
 248                         f, err := os.OpenFile(*outputDir+"/"+seqname+".npy", os.O_CREATE|os.O_WRONLY, 0777)
 249                         if err != nil {
 250                                 lastErr.Store(err)
 251                                 return
 252                         }
 253                         defer f.Close()
 254                         // gonpy closes our writer and ignores errors. Give it a nopCloser so we can close f properly.
 255                         npw, err := gonpy.NewWriter(nopCloser{f})
 256                         if err != nil {
 257                                 lastErr.Store(err)
 258                                 return
 259                         }
 260                         npw.Shape = []int{len(names), len(pdis) * 2}
 261                         err = npw.WriteInt8(data)
 262                         if err != nil {
 263                                 lastErr.Store(err)
 264                                 return
 265                         }
 266                         err = f.Close()
 267                         if err != nil {
 268                                 lastErr.Store(err)
 269                                 return
 270                         }
 271                 }()
 272         }
 273         wg.Wait()
 274         if e, ok := lastErr.Load().(error); ok {
 275                 err = e
 276                 return 1
 277         }
 278
 279         chunksize := (len(tilelib.variant) + *chunks - 1) / *chunks
 280         for chunk := 0; chunk < *chunks; chunk++ {
 281                 log.Infof("preparing chunk %d of %d", chunk+1, *chunks)
 282                 tagstart := chunk * chunksize
 283                 tagend := tagstart + chunksize
 284                 if tagend > len(tilelib.variant) {
 285                         tagend = len(tilelib.variant)
 286                 }
 287                 out, rows, cols := cgs2array(tilelib, names, lowqual, dropTiles, tagstart, tagend)
 288
 289                 var npw *gonpy.NpyWriter
 290                 var output io.WriteCloser
 291                 fnm := *outputDir + "/matrix.npy"
 292                 if *chunks > 1 {
 293                         fnm = fmt.Sprintf("%s/matrix.%d.npy", *outputDir, chunk)
 294                 }
 295                 output, err = os.OpenFile(fnm, os.O_CREATE|os.O_WRONLY, 0777)
 296                 if err != nil {
 297                         return 1
 298                 }
 299                 defer output.Close()
 300                 bufw := bufio.NewWriter(output)
 301                 npw, err = gonpy.NewWriter(nopCloser{bufw})
 302                 if err != nil {
 303                         return 1
 304                 }
 305                 if *onehot {
 306                         log.Info("recoding to onehot")
 307                         recoded, librefs, recodedcols := recodeOnehot(out, cols)
 308                         out, cols = recoded, recodedcols
 309                         if *librefsFilename != "" {
 310                                 log.Infof("writing onehot column mapping")
 311                                 err = cmd.writeLibRefs(*librefsFilename, tilelib, librefs)
 312                                 if err != nil {
 313                                         return 1
 314                                 }
 315                         }
 316                 }
 317                 log.WithFields(logrus.Fields{
 318                         "filename": fnm,
 319                         "rows":     rows,
 320                         "cols":     cols,
 321                 }).Info("writing numpy")
 322                 npw.Shape = []int{rows, cols}
 323                 npw.WriteInt16(out)
 324                 err = bufw.Flush()
 325                 if err != nil {
 326                         return 1
 327                 }
 328                 err = output.Close()
 329                 if err != nil {
 330                         return 1
 331                 }
 332         }
 333         return 0
 334 }
 335
 336 func (*exportNumpy) writeLibRefs(fnm string, tilelib *tileLibrary, librefs []tileLibRef) error {
 337         f, err := os.OpenFile(fnm, os.O_CREATE|os.O_WRONLY, 0666)
 338         if err != nil {
 339                 return err
 340         }
 341         defer f.Close()
 342         for i, libref := range librefs {
 343                 _, err = fmt.Fprintf(f, "%d,%d,%d\n", i, libref.Tag, libref.Variant)
 344                 if err != nil {
 345                         return err
 346                 }
 347         }
 348         return f.Close()
 349 }
 350
 351 func cgnames(tilelib *tileLibrary) (cgnames []string) {
 352         for name := range tilelib.compactGenomes {
 353                 cgnames = append(cgnames, name)
 354         }
 355         sort.Slice(cgnames, func(i, j int) bool {
 356                 return trimFilenameForLabel(cgnames[i]) < trimFilenameForLabel(cgnames[j])
 357         })
 358         return
 359 }
 360
 361 func lowqual(tilelib *tileLibrary) (lowqual []map[tileVariantID]bool) {
 362         lowqual = make([]map[tileVariantID]bool, len(tilelib.variant))
 363         for tag, variants := range tilelib.variant {
 364                 lq := lowqual[tag]
 365                 for varidx, hash := range variants {
 366                         if len(tilelib.hashSequence(hash)) == 0 {
 367                                 if lq == nil {
 368                                         lq = map[tileVariantID]bool{}
 369                                         lowqual[tag] = lq
 370                                 }
 371                                 lq[tileVariantID(varidx+1)] = true
 372                         }
 373                 }
 374         }
 375         return
 376 }
 377
 378 func cgs2array(tilelib *tileLibrary, names []string, lowqual []map[tileVariantID]bool, dropTiles []bool, tagstart, tagend int) (data []int16, rows, cols int) {
 379         rows = len(tilelib.compactGenomes)
 380         for tag := tagstart; tag < tagend; tag++ {
 381                 if len(dropTiles) <= tag || !dropTiles[tag] {
 382                         cols += 2
 383                 }
 384         }
 385         data = make([]int16, rows*cols)
 386         for row, name := range names {
 387                 cg := tilelib.compactGenomes[name]
 388                 outidx := 0
 389                 for tag := tagstart; tag < tagend && tag*2+1 < len(cg); tag++ {
 390                         if len(dropTiles) > tag && dropTiles[tag] {
 391                                 continue
 392                         }
 393                         for phase := 0; phase < 2; phase++ {
 394                                 v := cg[tag*2+phase]
 395                                 if v > 0 && lowqual[tag][v] {
 396                                         data[row*cols+outidx] = -1
 397                                 } else {
 398                                         data[row*cols+outidx] = int16(v)
 399                                 }
 400                                 outidx++
 401                         }
 402                 }
 403         }
 404         return
 405 }
 406
 407 func makeMask(regionsFilename string, expandRegions int) (*mask, error) {
 408         log.Printf("makeMask: reading %s", regionsFilename)
 409         rfile, err := zopen(regionsFilename)
 410         if err != nil {
 411                 return nil, err
 412         }
 413         defer rfile.Close()
 414         regions, err := io.ReadAll(rfile)
 415         if err != nil {
 416                 return nil, err
 417         }
 418
 419         log.Print("makeMask: building mask")
 420         var mask mask
 421         for _, line := range bytes.Split(regions, []byte{'\n'}) {
 422                 if bytes.HasPrefix(line, []byte{'#'}) {
 423                         continue
 424                 }
 425                 fields := bytes.Split(line, []byte{'\t'})
 426                 if len(fields) < 3 {
 427                         continue
 428                 }
 429                 refseqname := string(fields[0])
 430                 if strings.HasPrefix(refseqname, "chr") {
 431                         refseqname = refseqname[3:]
 432                 }
 433                 start, err1 := strconv.Atoi(string(fields[1]))
 434                 end, err2 := strconv.Atoi(string(fields[2]))
 435                 if err1 == nil && err2 == nil {
 436                         // BED
 437                 } else {
 438                         start, err1 = strconv.Atoi(string(fields[3]))
 439                         end, err2 = strconv.Atoi(string(fields[4]))
 440                         if err1 == nil && err2 == nil {
 441                                 // GFF/GTF
 442                                 end++
 443                         } else {
 444                                 return nil, fmt.Errorf("cannot parse input line as BED or GFF/GTF: %q", line)
 445                         }
 446                 }
 447                 mask.Add(refseqname, start-expandRegions, end+expandRegions)
 448         }
 449         log.Print("makeMask: mask.Freeze")
 450         mask.Freeze()
 451         return &mask, nil
 452 }
 453
 454 func chooseTiles(tilelib *tileLibrary, regionsFilename string, expandRegions int) (drop []bool, err error) {
 455         if regionsFilename == "" {
 456                 return
 457         }
 458         mask, err := makeMask(regionsFilename, expandRegions)
 459         if err != nil {
 460                 return
 461         }
 462
 463         tagset := tilelib.taglib.Tags()
 464         if len(tagset) == 0 {
 465                 err = errors.New("cannot choose tiles by region in a library without tags")
 466                 return
 467         }
 468         taglen := len(tagset[0])
 469
 470         log.Print("chooseTiles: check ref tiles")
 471         // Find position+size of each reference tile, and if it
 472         // intersects any of the desired regions, set drop[tag]=false.
 473         //
 474         // (Note it doesn't quite work to do the more obvious thing --
 475         // start with drop=false and change to true when ref tiles
 476         // intersect target regions -- because that would give us
 477         // drop=false for tiles that don't appear at all in the
 478         // reference.)
 479         //
 480         // TODO: (optionally?) don't drop tags for which some tile
 481         // variants are spanning tiles, i.e., where the reference tile
 482         // does not intersect the desired regions, but a spanning tile
 483         // from a genome does.
 484         drop = make([]bool, len(tilelib.variant))
 485         for i := range drop {
 486                 drop[i] = true
 487         }
 488         for refname, refseqs := range tilelib.refseqs {
 489                 for refseqname, reftiles := range refseqs {
 490                         if strings.HasPrefix(refseqname, "chr") {
 491                                 refseqname = refseqname[3:]
 492                         }
 493                         tileend := 0
 494                         for _, libref := range reftiles {
 495                                 if libref.Variant < 1 {
 496                                         err = fmt.Errorf("reference %q seq %q uses variant zero at tag %d", refname, refseqname, libref.Tag)
 497                                         return
 498                                 }
 499                                 seq := tilelib.TileVariantSequence(libref)
 500                                 if len(seq) < taglen {
 501                                         err = fmt.Errorf("reference %q seq %q uses tile %d variant %d with sequence len %d < taglen %d", refname, refseqname, libref.Tag, libref.Variant, len(seq), taglen)
 502                                         return
 503                                 }
 504                                 tilestart := tileend
 505                                 tileend = tilestart + len(seq) - taglen
 506                                 if mask.Check(refseqname, tilestart, tileend) {
 507                                         drop[libref.Tag] = false
 508                                 }
 509                         }
 510                 }
 511         }
 512
 513         log.Print("chooseTiles: done")
 514         return
 515 }
 516
 517 func recodeOnehot(in []int16, incols int) (out []int16, librefs []tileLibRef, outcols int) {
 518         rows := len(in) / incols
 519         maxvalue := make([]int16, incols)
 520         for row := 0; row < rows; row++ {
 521                 for col := 0; col < incols; col++ {
 522                         if v := in[row*incols+col]; maxvalue[col] < v {
 523                                 maxvalue[col] = v
 524                         }
 525                 }
 526         }
 527         outcol := make([]int, incols)
 528         dropped := 0
 529         for incol, maxv := range maxvalue {
 530                 outcol[incol] = outcols
 531                 if maxv == 0 {
 532                         dropped++
 533                 }
 534                 for v := 1; v <= int(maxv); v++ {
 535                         librefs = append(librefs, tileLibRef{Tag: tagID(incol), Variant: tileVariantID(v)})
 536                         outcols++
 537                 }
 538         }
 539         log.Printf("recodeOnehot: dropped %d input cols with zero maxvalue", dropped)
 540
 541         out = make([]int16, rows*outcols)
 542         for inidx, row := 0, 0; row < rows; row++ {
 543                 outrow := out[row*outcols:]
 544                 for col := 0; col < incols; col++ {
 545                         if v := in[inidx]; v > 0 {
 546                                 outrow[outcol[col]+int(v)-1] = 1
 547                         }
 548                         inidx++
 549                 }
 550         }
 551         return
 552 }
 553
 554 type nopCloser struct {
 555         io.Writer
 556 }
 557
 558 func (nopCloser) Close() error { return nil }
 559
 560 func trimFilenameForLabel(s string) string {
 561         if i := strings.LastIndex(s, "/"); i >= 0 {
 562                 s = s[i+1:]
 563         }
 564         s = strings.TrimSuffix(s, ".gz")
 565         s = strings.TrimSuffix(s, ".fa")
 566         s = strings.TrimSuffix(s, ".fasta")
 567         s = strings.TrimSuffix(s, ".1")
 568         s = strings.TrimSuffix(s, ".2")
 569         s = strings.TrimSuffix(s, ".gz")
 570         s = strings.TrimSuffix(s, ".vcf")
 571         return s
 572 }