action items 8 Jan 2025

[rvosa]

• debug early stop codons issue, e.g. in BGENL047-23, BGENL274-23, BGENL449-23, BGENL455-23, BGENL581-23, BGENL641-23, BGENL728-24, BGENL1370-24
• debug BLAST discrepancies as per email thread with Dan and Ben
• re-parameterize triage.py script to avoid all ambiguities

[SchistoDan]

Some samples with possible BLAST discrepencies to look into (not exhaustive - see XE-4013_barcode_fails-BV_output_verification.xlsx sheet 2 for ~100 possible discrepencies):
BSNHM1011-24_r_1_s_50_BSNHM1011-24
BSNHM1422-24_r_1_s_50_BSNHM1422-24
BSNHM1423-24_r_1_s_50_BSNHM1423-24
BSNHM1424-24_r_1_s_50_BSNHM1424-24
BSNHM1425-24_r_1_s_50_BSNHM1425-24
BSNHM1440-24_r_1_s_50_BSNHM1440-24
BSNHM1443-24_r_1_s_50_BSNHM1443-24
BSNTN2664-24_r_1_s_100_BSNTN2664-24_fastp

[rvosa]

At least one thing to consider is the BLAST parameterisation: the current config.yml has e-value threshold 1e-5, which is more stringent than web blast default. Also, the config looks at only 10 target seqs, while web blast considers 100, meaning a greater set of hits among which might be the expected family for reverse taxonomy.

(Historical note: the more stringent cutoff and the smaller result set was done to speed things up. A holistic analysis of the effect of the parameters might also look at running time.)