VcfStats.scala 18.6 KB
Newer Older
Peter van 't Hof's avatar
Peter van 't Hof committed
1
package nl.lumc.sasc.biopet.tools.vcfstats
2

3
import java.io.{File, FileOutputStream, IOException, PrintWriter}
4
import java.net.URLClassLoader
5

6
import htsjdk.variant.variantcontext.{Genotype, VariantContext}
7
import htsjdk.variant.vcf.VCFFileReader
8
import nl.lumc.sasc.biopet.utils.intervals.{BedRecord, BedRecordList}
9
import nl.lumc.sasc.biopet.utils.{ToolCommand, VcfUtils}
10
import org.apache.spark.{SparkConf, SparkContext}
Peter van 't Hof's avatar
Peter van 't Hof committed
11

12 13
import scala.collection.JavaConversions._
import scala.collection.mutable
14
import scala.sys.process.{Process, ProcessLogger}
Peter van 't Hof's avatar
Peter van 't Hof committed
15

16
/**
17
  * Created by pjvanthof on 14/07/2017.
18
  */
19
object VcfStats extends ToolCommand {
20

21
  type Args = VcfStatsArgs
22

23
  def main(args: Array[String]): Unit = {
24

25
    logger.info("Started")
26 27
    val argsParser = new VcfStatsOptParser(commandName)
    val cmdArgs = argsParser.parse(args, VcfStatsArgs()) getOrElse (throw new IllegalArgumentException)
28

Peter van 't Hof's avatar
Peter van 't Hof committed
29
    val reader = new VCFFileReader(cmdArgs.inputFile, true)
30 31 32
    val header = reader.getFileHeader
    val samples = header.getSampleNamesInOrder.toList

Peter van 't Hof's avatar
Peter van 't Hof committed
33 34
    reader.close()

Peter van 't Hof's avatar
Peter van 't Hof committed
35
    val adInfoTags = {
36 37 38
      (for (infoTag <- cmdArgs.infoTags if !defaultInfoFields.contains(infoTag)) yield {
        require(header.getInfoHeaderLine(infoTag) != null,
                "Info tag '" + infoTag + "' not found in header of vcf file")
Peter van 't Hof's avatar
Peter van 't Hof committed
39
        infoTag
40 41 42
      }) ::: (for (line <- header.getInfoHeaderLines if cmdArgs.allInfoTags
                   if !defaultInfoFields.contains(line.getID)
                   if !cmdArgs.infoTags.contains(line.getID)) yield {
Peter van 't Hof's avatar
Peter van 't Hof committed
43 44 45
        line.getID
      }).toList ::: defaultInfoFields
    }
46

47 48 49 50
    val adGenotypeTags = (for (genotypeTag <- cmdArgs.genotypeTags
                               if !defaultGenotypeFields.contains(genotypeTag)) yield {
      require(header.getFormatHeaderLine(genotypeTag) != null,
              "Format tag '" + genotypeTag + "' not found in header of vcf file")
51
      genotypeTag
52 53 54 55
    }) ::: (for (line <- header.getFormatHeaderLines if cmdArgs.allGenotypeTags
                 if !defaultGenotypeFields.contains(line.getID)
                 if !cmdArgs.genotypeTags.contains(line.getID)
                 if line.getID != "PL") yield {
56
      line.getID
Peter van 't Hof's avatar
Peter van 't Hof committed
57
    }).toList ::: defaultGenotypeFields
58

59 60 61 62 63 64 65 66 67 68 69 70 71 72
    logger.info("Init spark context")

    val jars = ClassLoader.getSystemClassLoader
      .asInstanceOf[URLClassLoader]
      .getURLs
      .map(_.getFile)
    val conf = new SparkConf()
      .setAppName(commandName)
      .setMaster(cmdArgs.sparkMaster.getOrElse(s"local[${cmdArgs.localThreads}]"))
      .setJars(jars)
    val sc = new SparkContext(conf)
    logger.info("Spark context is up")

    val regions = (cmdArgs.intervals match {
Peter van 't Hof's avatar
Peter van 't Hof committed
73 74
      case Some(i) =>
        BedRecordList.fromFile(i).validateContigs(cmdArgs.referenceFile)
75
      case _ => BedRecordList.fromReference(cmdArgs.referenceFile)
76 77 78
    }).combineOverlap
      .scatter(cmdArgs.binSize, maxContigsInSingleJob = Some(cmdArgs.maxContigsInSingleJob))
    val contigs = regions.flatMap(_.map(_.chr)).distinct
Peter van 't Hof's avatar
Peter van 't Hof committed
79

80 81 82
    val regionStats = sc
      .parallelize(regions, regions.size)
      .map(readBin(_, samples, cmdArgs, adInfoTags, adGenotypeTags))
Peter van 't Hof's avatar
Peter van 't Hof committed
83

84
    val totalStats = regionStats.reduce(_ += _)
85

86
    totalStats.writeToFile(new File(cmdArgs.outputDir, "stats.json"), samples, adGenotypeTags, adInfoTags, sampleDistributions)
Peter van 't Hof's avatar
Peter van 't Hof committed
87

88
    writeOverlap(totalStats,
89 90 91
                 _.genotypeOverlap,
                 cmdArgs.outputDir + "/sample_compare/genotype_overlap",
                 samples)
92
    writeOverlap(totalStats,
93 94 95
                 _.alleleOverlap,
                 cmdArgs.outputDir + "/sample_compare/allele_overlap",
                 samples)
96

97
    sc.stop
98
    logger.info("Done")
Peter van 't Hof's avatar
Peter van 't Hof committed
99 100
  }

101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136
  def readBin(bedRecords: List[BedRecord],
              samples: List[String],
              cmdArgs: Args,
              adInfoTags: List[String],
              adGenotypeTags: List[String]): Stats = {
    val reader = new VCFFileReader(cmdArgs.inputFile, true)
    val stats = Stats.emptyStats(samples)

    for (bedRecord <- bedRecords) {
      logger.info("Starting on: " + bedRecord)

      val samInterval = bedRecord.toSamInterval

      val query =
        reader.query(samInterval.getContig, samInterval.getStart, samInterval.getEnd)
      if (!query.hasNext) {
        Stats.mergeNestedStatsMap(stats.generalStats, fillGeneral(adInfoTags))
        for (sample <- samples) yield {
          Stats.mergeNestedStatsMap(stats.samplesStats(sample).genotypeStats,
                                    fillGenotype(adGenotypeTags))
        }
      }

      for (record <- query if record.getStart <= samInterval.getEnd) {
        Stats.mergeNestedStatsMap(stats.generalStats, checkGeneral(record, adInfoTags))
        for (sample1 <- samples) yield {
          val genotype = record.getGenotype(sample1)
          Stats.mergeNestedStatsMap(stats.samplesStats(sample1).genotypeStats,
                                    checkGenotype(record, genotype, adGenotypeTags))
          for (sample2 <- samples) {
            val genotype2 = record.getGenotype(sample2)
            if (genotype.getAlleles == genotype2.getAlleles)
              stats.samplesStats(sample1).sampleToSample(sample2).genotypeOverlap += 1
            stats.samplesStats(sample1).sampleToSample(sample2).alleleOverlap += VcfUtils
              .alleleOverlap(genotype.getAlleles.toList, genotype2.getAlleles.toList)
          }
Peter van 't Hof's avatar
Peter van 't Hof committed
137
        }
Peter van 't Hof's avatar
Peter van 't Hof committed
138 139
      }
    }
140 141 142
    reader.close()

    stats
Peter van 't Hof's avatar
Peter van 't Hof committed
143 144
  }

145 146 147 148
  val defaultGenotypeFields =
    List("DP", "GQ", "AD", "AD-ref", "AD-alt", "AD-used", "AD-not_used", "general")

  val defaultInfoFields = List("QUAL", "general", "AC", "AF", "AN", "DP")
Peter van 't Hof's avatar
Peter van 't Hof committed
149

150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191
  val sampleDistributions = List("Het",
                                 "HetNonRef",
                                 "Hom",
                                 "HomRef",
                                 "HomVar",
                                 "Mixed",
                                 "NoCall",
                                 "NonInformative",
                                 "Available",
                                 "Called",
                                 "Filtered",
                                 "Variant")

  /** Function to write sample to sample compare tsv's / heatmaps */
  def writeOverlap(stats: Stats,
                   function: SampleToSampleStats => Int,
                   prefix: String,
                   samples: List[String],
                   plots: Boolean = true): Unit = {
    val absFile = new File(prefix + ".abs.tsv")
    val relFile = new File(prefix + ".rel.tsv")

    absFile.getParentFile.mkdirs()

    val absWriter = new PrintWriter(absFile)
    val relWriter = new PrintWriter(relFile)

    absWriter.println(samples.mkString("\t", "\t", ""))
    relWriter.println(samples.mkString("\t", "\t", ""))
    for (sample1 <- samples) {
      val values = for (sample2 <- samples)
        yield function(stats.samplesStats(sample1).sampleToSample(sample2))

      absWriter.println(values.mkString(sample1 + "\t", "\t", ""))

      val total = function(stats.samplesStats(sample1).sampleToSample(sample1))
      relWriter.println(values.map(_.toFloat / total).mkString(sample1 + "\t", "\t", ""))
    }
    absWriter.close()
    relWriter.close()

    if (plots) plotHeatmap(relFile)
192 193
  }

194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236
  /** Plots heatmaps on target tsv file */
  def plotHeatmap(file: File) {
    executeRscript(
      "plotHeatmap.R",
      Array(
        file.getAbsolutePath,
        file.getAbsolutePath.stripSuffix(".tsv") + ".heatmap.png",
        file.getAbsolutePath.stripSuffix(".tsv") + ".heatmap.clustering.png",
        file.getAbsolutePath.stripSuffix(".tsv") + ".heatmap.dendrogram.png"
      )
    )
  }

  /** Function to execute Rscript as subproces */
  def executeRscript(resource: String, args: Array[String]): Unit = {
    val is = getClass.getResourceAsStream(resource)
    val file = File.createTempFile("script.", "." + resource)
    file.deleteOnExit()
    val os = new FileOutputStream(file)
    org.apache.commons.io.IOUtils.copy(is, os)
    os.close()

    val command: String = "Rscript " + file + " " + args.mkString(" ")

    logger.info("Starting: " + command)
    try {
      val process = Process(command).run(ProcessLogger(x => logger.debug(x), x => logger.debug(x)))
      if (process.exitValue() == 0) logger.info("Done: " + command)
      else {
        logger.warn("Failed: " + command)
        if (!logger.isDebugEnabled) logger.warn("Use -l debug for more info")
      }
    } catch {
      case e: IOException =>
        logger.warn("Failed: " + command)
        logger.debug(e)
    }
  }

  /** Function to check sample/genotype specific stats */
  protected[tools] def checkGenotype(
      record: VariantContext,
      genotype: Genotype,
237
      additionalTags: List[String]): Map[String, Map[Any, Int]] = {
238 239 240 241 242 243 244 245
    val buffer = mutable.Map[String, Map[Any, Int]]()

    def addToBuffer(key: String, value: Any, found: Boolean): Unit = {
      val map = buffer.getOrElse(key, Map())
      if (found) buffer += key -> (map + (value -> (map.getOrElse(value, 0) + 1)))
      else buffer += key -> (map + (value -> map.getOrElse(value, 0)))
    }

246 247
    buffer += "DP" -> Map((if (genotype.hasDP) genotype.getDP else "not set") -> 1)
    buffer += "GQ" -> Map((if (genotype.hasGQ) genotype.getGQ else "not set") -> 1)
248

249 250
    val usedAlleles =
      (for (allele <- genotype.getAlleles) yield record.getAlleleIndex(allele)).toList
251

252 253 254 255 256 257 258 259 260 261 262 263 264
    addToBuffer("general", "Total", found = true)
    addToBuffer("general", "Het", genotype.isHet)
    addToBuffer("general", "HetNonRef", genotype.isHetNonRef)
    addToBuffer("general", "Hom", genotype.isHom)
    addToBuffer("general", "HomRef", genotype.isHomRef)
    addToBuffer("general", "HomVar", genotype.isHomVar)
    addToBuffer("general", "Mixed", genotype.isMixed)
    addToBuffer("general", "NoCall", genotype.isNoCall)
    addToBuffer("general", "NonInformative", genotype.isNonInformative)
    addToBuffer("general", "Available", genotype.isAvailable)
    addToBuffer("general", "Called", genotype.isCalled)
    addToBuffer("general", "Filtered", genotype.isFiltered)
    addToBuffer("general", "Variant", genotype.isHetNonRef || genotype.isHet || genotype.isHomVar)
265

266 267 268 269 270 271 272 273 274 275 276 277
    if (genotype.hasAD) {
      val ad = genotype.getAD
      for (i <- 0 until ad.size if ad(i) > 0) {
        addToBuffer("AD", ad(i), found = true)
        if (i == 0) addToBuffer("AD-ref", ad(i), found = true)
        if (i > 0) addToBuffer("AD-alt", ad(i), found = true)
        if (usedAlleles.contains(i)) addToBuffer("AD-used", ad(i), found = true)
        else addToBuffer("AD-not_used", ad(i), found = true)
      }
    }

    val skipTags = List("DP", "GQ", "AD", "AD-ref", "AD-alt", "AD-used", "AD-not_used", "general")
278 279

    for (tag <- additionalTags if !skipTags.contains(tag)) {
280 281 282
      val value = genotype.getAnyAttribute(tag)
      if (value == null) addToBuffer(tag, "notset", found = true)
      else addToBuffer(tag, value, found = true)
283 284
    }

285
    buffer.toMap
286 287
  }

288
  /** Function to check all general stats, all info expect sample/genotype specific stats */
289 290
  protected[tools] def checkGeneral(
      record: VariantContext,
291
      additionalTags: List[String]): Map[String, Map[Any, Int]] = {
Peter van 't Hof's avatar
Peter van 't Hof committed
292 293
    val buffer = mutable.Map[String, Map[Any, Int]]()

294
    def addToBuffer(key: String, value: Any, found: Boolean): Unit = {
Peter van 't Hof's avatar
Peter van 't Hof committed
295
      val map = buffer.getOrElse(key, Map())
296
      if (found) buffer += key -> (map + (value -> (map.getOrElse(value, 0) + 1)))
Peter van 't Hof's avatar
Peter van 't Hof committed
297
      else buffer += key -> (map + (value -> map.getOrElse(value, 0)))
Peter van 't Hof's avatar
Peter van 't Hof committed
298 299
    }

Peter van 't Hof's avatar
Peter van 't Hof committed
300
    addToBuffer("QUAL", Math.round(record.getPhredScaledQual), found = true)
Peter van 't Hof's avatar
Peter van 't Hof committed
301

302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338
    addToBuffer("SampleDistribution-Het",
                record.getGenotypes.count(genotype => genotype.isHet),
                found = true)
    addToBuffer("SampleDistribution-HetNonRef",
                record.getGenotypes.count(genotype => genotype.isHetNonRef),
                found = true)
    addToBuffer("SampleDistribution-Hom",
                record.getGenotypes.count(genotype => genotype.isHom),
                found = true)
    addToBuffer("SampleDistribution-HomRef",
                record.getGenotypes.count(genotype => genotype.isHomRef),
                found = true)
    addToBuffer("SampleDistribution-HomVar",
                record.getGenotypes.count(genotype => genotype.isHomVar),
                found = true)
    addToBuffer("SampleDistribution-Mixed",
                record.getGenotypes.count(genotype => genotype.isMixed),
                found = true)
    addToBuffer("SampleDistribution-NoCall",
                record.getGenotypes.count(genotype => genotype.isNoCall),
                found = true)
    addToBuffer("SampleDistribution-NonInformative",
                record.getGenotypes.count(genotype => genotype.isNonInformative),
                found = true)
    addToBuffer("SampleDistribution-Available",
                record.getGenotypes.count(genotype => genotype.isAvailable),
                found = true)
    addToBuffer("SampleDistribution-Called",
                record.getGenotypes.count(genotype => genotype.isCalled),
                found = true)
    addToBuffer("SampleDistribution-Filtered",
                record.getGenotypes.count(genotype => genotype.isFiltered),
                found = true)
    addToBuffer("SampleDistribution-Variant",
                record.getGenotypes.count(genotype =>
                  genotype.isHetNonRef || genotype.isHet || genotype.isHomVar),
                found = true)
Peter van 't Hof's avatar
Peter van 't Hof committed
339

Peter van 't Hof's avatar
Peter van 't Hof committed
340
    addToBuffer("general", "Total", found = true)
341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359
    addToBuffer("general", "Biallelic", record.isBiallelic)
    addToBuffer("general", "ComplexIndel", record.isComplexIndel)
    addToBuffer("general", "Filtered", record.isFiltered)
    addToBuffer("general", "FullyDecoded", record.isFullyDecoded)
    addToBuffer("general", "Indel", record.isIndel)
    addToBuffer("general", "Mixed", record.isMixed)
    addToBuffer("general", "MNP", record.isMNP)
    addToBuffer("general", "MonomorphicInSamples", record.isMonomorphicInSamples)
    addToBuffer("general", "NotFiltered", record.isNotFiltered)
    addToBuffer("general", "PointEvent", record.isPointEvent)
    addToBuffer("general", "PolymorphicInSamples", record.isPolymorphicInSamples)
    addToBuffer("general", "SimpleDeletion", record.isSimpleDeletion)
    addToBuffer("general", "SimpleInsertion", record.isSimpleInsertion)
    addToBuffer("general", "SNP", record.isSNP)
    addToBuffer("general", "StructuralIndel", record.isStructuralIndel)
    addToBuffer("general", "Symbolic", record.isSymbolic)
    addToBuffer("general", "SymbolicOrSV", record.isSymbolicOrSV)
    addToBuffer("general", "Variant", record.isVariant)

Peter van 't Hof's avatar
Peter van 't Hof committed
360 361 362
    val skipTags = List("QUAL", "general")

    for (tag <- additionalTags if !skipTags.contains(tag)) {
363
      val value = record.getAttribute(tag)
Peter van 't Hof's avatar
Peter van 't Hof committed
364 365
      if (value == null) addToBuffer(tag, "notset", found = true)
      else addToBuffer(tag, value, found = true)
366 367
    }

368
    buffer.toMap
Peter van 't Hof's avatar
Peter van 't Hof committed
369 370
  }

371
  protected[tools] def fillGeneral(
372
      additionalTags: List[String]): Map[String, Map[Any, Int]] = {
373 374 375 376 377 378 379 380
    val buffer = mutable.Map[String, Map[Any, Int]]()

    def addToBuffer(key: String, value: Any, found: Boolean): Unit = {
      val map = buffer.getOrElse(key, Map())
      if (found) buffer += key -> (map + (value -> (map.getOrElse(value, 0) + 1)))
      else buffer += key -> (map + (value -> map.getOrElse(value, 0)))
    }

381 382 383 384 385 386 387 388 389 390 391 392 393 394
    addToBuffer("QUAL", "not set", found = false)

    addToBuffer("SampleDistribution-Het", "not set", found = false)
    addToBuffer("SampleDistribution-HetNonRef", "not set", found = false)
    addToBuffer("SampleDistribution-Hom", "not set", found = false)
    addToBuffer("SampleDistribution-HomRef", "not set", found = false)
    addToBuffer("SampleDistribution-HomVar", "not set", found = false)
    addToBuffer("SampleDistribution-Mixed", "not set", found = false)
    addToBuffer("SampleDistribution-NoCall", "not set", found = false)
    addToBuffer("SampleDistribution-NonInformative", "not set", found = false)
    addToBuffer("SampleDistribution-Available", "not set", found = false)
    addToBuffer("SampleDistribution-Called", "not set", found = false)
    addToBuffer("SampleDistribution-Filtered", "not set", found = false)
    addToBuffer("SampleDistribution-Variant", "not set", found = false)
Peter van 't Hof's avatar
Peter van 't Hof committed
395 396

    addToBuffer("general", "Total", found = false)
397 398
    addToBuffer("general", "Biallelic", found = false)
    addToBuffer("general", "ComplexIndel", found = false)
Peter van 't Hof's avatar
Peter van 't Hof committed
399
    addToBuffer("general", "Filtered", found = false)
400 401 402 403 404 405 406 407 408 409 410 411 412 413
    addToBuffer("general", "FullyDecoded", found = false)
    addToBuffer("general", "Indel", found = false)
    addToBuffer("general", "Mixed", found = false)
    addToBuffer("general", "MNP", found = false)
    addToBuffer("general", "MonomorphicInSamples", found = false)
    addToBuffer("general", "NotFiltered", found = false)
    addToBuffer("general", "PointEvent", found = false)
    addToBuffer("general", "PolymorphicInSamples", found = false)
    addToBuffer("general", "SimpleDeletion", found = false)
    addToBuffer("general", "SimpleInsertion", found = false)
    addToBuffer("general", "SNP", found = false)
    addToBuffer("general", "StructuralIndel", found = false)
    addToBuffer("general", "Symbolic", found = false)
    addToBuffer("general", "SymbolicOrSV", found = false)
Peter van 't Hof's avatar
Peter van 't Hof committed
414
    addToBuffer("general", "Variant", found = false)
415

416
    val skipTags = List("QUAL", "general")
417 418

    for (tag <- additionalTags if !skipTags.contains(tag)) {
419
      addToBuffer(tag, "not set", found = false)
420 421
    }

422
    buffer.toMap
423 424
  }

425
  protected[tools] def fillGenotype(
426
      additionalTags: List[String]): Map[String, Map[Any, Int]] = {
Peter van 't Hof's avatar
Peter van 't Hof committed
427
    val buffer = mutable.Map[String, Map[Any, Int]]()
Peter van 't Hof's avatar
Peter van 't Hof committed
428

429
    def addToBuffer(key: String, value: Any, found: Boolean): Unit = {
Peter van 't Hof's avatar
Peter van 't Hof committed
430
      val map = buffer.getOrElse(key, Map())
431
      if (found) buffer += key -> (map + (value -> (map.getOrElse(value, 0) + 1)))
Peter van 't Hof's avatar
Peter van 't Hof committed
432
      else buffer += key -> (map + (value -> map.getOrElse(value, 0)))
Peter van 't Hof's avatar
Peter van 't Hof committed
433 434
    }

435 436
    addToBuffer("DP", "not set", found = false)
    addToBuffer("GQ", "not set", found = false)
Peter van 't Hof's avatar
Peter van 't Hof committed
437

438 439 440 441 442 443 444 445 446 447 448 449 450
    addToBuffer("general", "Total", found = false)
    addToBuffer("general", "Het", found = false)
    addToBuffer("general", "HetNonRef", found = false)
    addToBuffer("general", "Hom", found = false)
    addToBuffer("general", "HomRef", found = false)
    addToBuffer("general", "HomVar", found = false)
    addToBuffer("general", "Mixed", found = false)
    addToBuffer("general", "NoCall", found = false)
    addToBuffer("general", "NonInformative", found = false)
    addToBuffer("general", "Available", found = false)
    addToBuffer("general", "Called", found = false)
    addToBuffer("general", "Filtered", found = false)
    addToBuffer("general", "Variant", found = false)
Peter van 't Hof's avatar
Peter van 't Hof committed
451

Peter van 't Hof's avatar
Peter van 't Hof committed
452 453 454
    val skipTags = List("DP", "GQ", "AD", "AD-ref", "AD-alt", "AD-used", "AD-not_used", "general")

    for (tag <- additionalTags if !skipTags.contains(tag)) {
455
      addToBuffer(tag, 0, found = false)
456
    }
Peter van 't Hof's avatar
Peter van 't Hof committed
457

458
    buffer.toMap
459 460
  }
}