Known issues

P value conversion during extracting lead variants

gwaslab<=3.4.18 : In older versions, gwaslab converts MLOG10P to P (if P column is not available) and then uses P to extract lead variants using .get_lead(). But during conversion, since MLOG10P was set as float32, and then P was accordingly set as float32. The minimum of float32 is around 1e-38, below which P values will become 0. This could cause error when extracting lead variants.
Solution: simply use scaled=True to directly use MLOG10P instead of P values.

gwaslab<=3.4.22 : When extracting lead variants using .get_lead(), if the lead variant is the last variant in all significant variants, it will be mistakenly skipped.
Solution: Update to new version of gwaslab.

gwaslab<=3.4.39 : the color assigned to each variant is actually the color for the lower LD r2 category. For example, variants with LD>0.8 will be colored with the color for 0.8>LD>0.6.
Solution: Update to new version of gwaslab.

3.4.43 <= gwaslab <= 3.5.7 : If the coordinates in sumstats are not unsorted, the alignment may fail.
Solution: Update to new version of gwaslab.

A .tbi or .csi that was not built from the local VCF can be inconsistent with it. bcftools view -r 14 may start mid-chromosome (for example around 50 Mb) even though a direct interval query such as 14:30000400 still returns the variant. Sweep-mode .assign_rsid2() used whole-chromosome -r CHR with -T targets, so those prefix variants looked absent from the reference VCF. Downloaded dbSNP v151 VCF/index pairs are one known example.
This is not a GWASLab bug in the VCF itself. Rebuild the index from the local VCF (bcftools index -t -f <vcf>), then re-run assignment. Current GWASLab queries -r CHR:minPOS-maxPOS and warns under these conditions (see Assigning rsID and Download reference data):
- Lookup stream (sweep extract, n_targets > 0): no records returned for that chromosome, or first returned POS is after the smallest target POS and > 5% of that chromosome’s targets sit before the first returned record.
- Assignment-rate outlier (after VCF/sweep fill): among chromosomes with ≥ 500 eligible variants (need ≥ 5 such chromosomes), fill rate is ≥ 0.15 below the median (median_rate − rate ≥ 0.15).