10x-sc-5gex with BAM input

[racng]

If I use sample_alignments.bam output from running cellranger multi as input for the 10x-sc-5gex preset, does mixcr use the corrected cell barcodes and UMI stored in the BAM tags? I am curious to try using BAM as the input, because this review here indicates that using BAM instead of FASTQ input yields better sensitivity and accuracy.

[mizraelson]

By default, MiXCR will still run the error correction steps regardless of whether the input file is a BAM or a FASTQ, even if the barcodes were corrected beforehand (MiXCR cannot tell if the barcodes were pre-corrected, as it only sees the sequences). The difference probably depends on how you calculate sensitivity and accuracy, and probably on the dataset (number of cell barcodes, number of UMIs, sequencing depth). We haven’t found any drawbacks of using FASTA compared to BAM.