influence of BED file on variant results

[djomlaart]

Hi, we were wondering what are the effects of specifying BED file on variant calling process.

Namely we’ve compared two different scenarios - specifying BED file as a method to restrict the variants calls on decoy/alt regions vs pre-filtering the BAM file (with the same BED file) prior to running Manta. We got a slightly different results. Initially we’ve attributed this to different number of reads being available to manta in order to calculate fragment size statistics.

But in the second scenario we’ve repeated calling on pre-filtered BAM file with the addition of BED file provided to Manta as input (although there were no-reads outside the specified regions since the same BED was used for pre-filtering BAM and for Manta). We’ve noticed significant less difference in variant calls when comparing these to Manta ran on full WGS with provided BED file.

Does this imply that by providing the BED file Manta treats the available data differently based wether BED file is present or not?