Searching thousands of genomes to classify somatic and novel structural variants using STIX

Abstract

Structural variants are associated with cancers and developmental disorders, but challenges with estimating population frequency remain a barrier to prioritizing mutations over inherited variants. In particular, variability in variant calling heuristics and filtering limits the use of current structural variant catalogs. We present STIX, a method that, instead of relying on variant calls, indexes and searches the raw alignments from thousands of samples to enable more comprehensive allele frequency estimation.

The benefit of a complete reference genome for cancer structural variant analysis

“…Although SV detection is improved (i.e. no FP SVs detected), using CHM13-T2T limits variant prioritization and interpretation as population frequencies are not available on this reference, and GRCh38 has more informative annotation databases (Collins et al 2020; Nicholas et al 2022; Chowdhury et al 2022). Thus, we used recent advances in liftover of alignments to take advantage of both the improved mapping using a CHM13-T2T reference (Chen et al 2024) and the years of annotation and curation of the GRCh38 reference genome.…”

“…Furthermore, this liftover approach does not double the analytical time nor the costs for analysis as it utilizes a chain file to rapidly liftover the read alignments without additional pairwise alignments needed (Chen et al 2024). This also permits downstream utilization of annotation databases and approaches such as gnomadSV (Collins et al 2021) and STIX (Chowdhury et al 2022) to further filter and improve variant prioritization.…”

Implications of noncoding regulatory functions in the development of insulinomas

“…Population SVs were inferred from 1,000G catalog. To this end 2,504 low-coverage BAMs were downloaded from the 1,000 genomes AWS S3 bucket (s3://1000genomes/phase3/data/) to build a control reference STIX database using excord (version 0.2.4) (https://github.com/brentp/excord), giggle 85 (version 0.6.3) and STIX 85 (version 1.0). First, SV alignment evidence was extracted from BAM using excord, considering both discordand and split reads (discordant distance=500).…”

“…Next indexes for each excord evidence were generated by giggle. Finally, the database was created using STIX as described elsewhere 85 . The same procedure was applied to the patient sample cohort to obtain an insulinoma STIX database.…”

“…SVs were called using Delly 85 (version 1.0). First, SV alignment evidence was extracted from BAM using excord, considering both discordand and split reads (discordant distance=500).…”