<#2391 [modules] Small test dataset with real SV s...
# test-data
g
#2391 [modules] Small test dataset with real SV signal (split/discordant reads) for GRIDSS modules Issue created by imsarath Description The current
modules
test data used by the nf-core GRIDSS modules (
gridss/gridss
,
gridss/preprocess
,
gridss/assemble
,
gridss/call
) has no structural variant signal:
test.paired_end.sorted.bam
against
genome.fasta
contains no split or discordant read pairs, and the snapshots record
variantCount=0
. GRIDSS only loads the BWA index when it has reads to realign. Without split/discordant reads, it never checks whether the index files exist. As a result, the module tests can pass even if the BWA index is staged incorrectly. We hit this downstream in genomic-medicine-sweden/autoseq: the index files were not symlinked properly, but nf-test still passed. Existing data checked • `genomics/homo_sapiens/illumina/bam/test.paired_end.sorted.bam`: used by all GRIDSS module tests; produces 0 SV calls. • `genomics/homo_sapiens/illumina/vcf/test.genome_21.somatic_sv.vcf.gz`: Manta output on the chr21 test BAMs; header only, 0 records. • `genomics/homo_sapiens/illumina/vcf/chr21/simulated_sv{,2}.vcf.gz`: SURVIVOR-simulated SV VCFs only; no matching reads. • `genomics/homo_sapiens/illumina/vcf/gridss.somatic.vcf`: whole-genome GRIDSS output; no matching reads. •
oncoanalyser
branch `sample_data/simulated_reads/wgts/markdups_bam/subject_a.{tumor,normal}`: has some split/discordant reads, but is aligned to full GRCh38, so it needs a full-size reference and BWA index. That makes it too large for module tests. Request Add a small paired tumor/normal (or single-sample) dataset to the
modules
branch with real SV evidence (split and discordant reads, e.g. a few deletions/inversions/translocations) that matches the existing small
genome.fasta
(or a small companion reference with a BWA index). With this data: • GRIDSS emits at least one breakpoint, so the snapshots can assert
variantCount > 0
. • The BWA index is actually used, so a missing or mis-staged index fails the test. Possible approach: simulate SVs into the existing small reference (e.g. SURVIVOR
simSV
+ `wgsim`/`art`), align with `bwa`/`bwa-mem2`, then sort and index. nf-core/test-datasets