|
|
Candidate Indel Search
Strelka scans through the genome using sequence alignments from the normal sample and tumor sample together to find a joint set of candidate indels. The information in sequence alignments is supplemented with externally generated candidate indels discovered by Manta. Manta provides external candidate indels to Strelka for indels of size 50 and below.
Candidate indels are used for realignment of reads, during which each candidate indel is evaluated as a potential somatic indel. Any other types of indels are considered noise indels. If a better alignment is not found, these indels are allowed to remain in the read alignments; otherwise, they are not used.
The candidate indel thresholds are designed so that the joint candidate indel set is at least the combined set found if the Small Variant Caller (Starling) is run on the individual samples. Specifically, where a minimum number of nominating reads is required for candidacy in Starling, Strelka requires the same minimum number of nominating reads from the combined input. Strelka requires that at least 1 sample contains a minimum fraction of supporting reads among the sample reads for candidacy.