Trying to extract per-index statistics from MiSeq's GenerateFASTQRunStatistics.xml file with XPath.

Question

[Charles Plessy]

MiSeq runs contain some report files in XML format. In particular, Data/Intensities/BaseCalls/Alignment/GenerateFASTQRunStatistics.xml indicates various numbers such as the count of reads passing filter, for the whole run and for each index separately. Here is an oversiplified example.

<StatisticsGenerateFASTQ>
  <RunStats>
    <NumberOfClustersPF>11433659</NumberOfClustersPF>
    <NumberOfClustersRaw>11969395</NumberOfClustersRaw>
  </RunStats>
  <OverallSamples>
    <SummarizedSampleStatistics>
      <NumberOfClustersPF>43181</NumberOfClustersPF>
      <NumberOfClustersRaw>49080</NumberOfClustersRaw>
      <SampleNumber>1</SampleNumber>
    </SummarizedSampleStatistics>
    <SummarizedSampleStatistics>
      <NumberOfClustersPF>79129</NumberOfClustersPF>
      <NumberOfClustersRaw>85016</NumberOfClustersRaw>
      <SampleNumber>2</SampleNumber>
    </SummarizedSampleStatistics>
  </OverallSamples>
  <PairedEndByGenome />
  <Samples />
</StatisticsGenerateFASTQ>

Using the XmlStarlet comand-line tool, I could extract the total number of clusters passing filter with the command xmlstarlet sel -t -v //RunStats/NumberOfClustersPF GenerateFASTQRunStatistics.xml. However, to do so for a given index, one needs to first identify which SummarizedSampleStatistics node contains a SampleNumber child that contains a given index number, and then extract the value from the NumberOfClustersPF sibling node.

Does anybody know an XPath expression doing that ? (so that it could become a one-liner with XmlStarlet). I tried things like //ancestor::SampleNumber[.="2"]/NumberOfClustersPF without success...