Alignment

1. 01Indexed random access through the .bai, so only the requested window is read
2. 02SAM text opens too, with no index — it is read whole
3. 03Coverage track over the current window
4. 04Read pileup drawn from the CIGAR, with insertions and deletions in place
5. 05Per-read readouts: MAPQ, mate flags, insert size and strand
6. 06Reference and locus selection, including search by locus
7. 07Long reads and short reads in the same viewer
8. 08A reference FASTA can be loaded: its bases show under the ruler and mismatches are called against it
9. 09Export the current window, or the whole file, as SAM with every optional field intact

BAM parsing is a port of @gmod/bam, exercised against the same corpora the upstream library tests use. The SAM reader and writer follow the SAMv1 spec and htslib output rules, and are checked byte-for-byte against real samtools output; the FASTA index is a port of samtools faidx, checked byte-for-byte against pyfaidx. The panel runs on @ports/bam.

Everything runs client-side. Files are parsed in the browser and projects are saved on this machine, so an unpublished sequence is never uploaded.

Do I need the .bai index file?

For a BAM, yes — random access to a locus is what the index provides, and it is the same requirement IGV and JBrowse have. A SAM needs none: it is text, so it is read whole.

Can I see the reference bases under the reads?

Yes, if the project has the reference FASTA. Pick it in the Reference selector and the bases appear under the ruler, with mismatches computed against the real sequence rather than the optional MD tag most aligners omit.

Is the BAM uploaded to a server?

No. The file is read in the browser, and only the byte ranges for the current window are decoded.