Genetics

Long-read genome sequencing adds ASD diagnoses by resolving complex SVs and repeats

In 63 autism spectrum disorder families (267 individuals), long-read WGS was integrated with short-read data for structural and repeat variant calling. Long-read sequencing increased detection of gene-disrupting SVs and tandem repeats by 33% and 38%, respectively, versus short-read alone. Previously unrecognized exonic de novo germline and somatic SVs, including complex nested duplication–deletion events, were identified. Phased genetics plus methylation revealed deletions of imprinted genes and methylation effects of intermediate FMR1 CGG repeat expansions. Rare SVs, tandem repeats, and damaging SNVs jointly explained 7.4% of ASD heritability, underscoring their diagnostic relevance. 1

Tumor HRD signatures support BRCA1/2 germline variant classification

Breast tumor and germline whole-genome data from 350 patients were analyzed for germline BRCA1/2 status and tumor HRD signatures. Somatic HR deficiency or proficiency significantly predicted the presence or absence of germline pathogenic BRCA1/2 variants. The CHORD algorithm, which distinguishes BRCA1- versus BRCA2-type HRD, provided gene-specific evidence reaching moderate pathogenic strength for the corresponding gene. HRD predictions were also applied to BRCA1/2 VUS, offering additional weighted evidence that could reduce germline classification uncertainty. 2

ML-ExonCNV improves rare exon-level CNV detection from WES

ML-ExonCNV uses an XGBoost multi-expert ensemble to detect rare exon-level CNVs in WES data. The model was trained on 22,364 qPCR-validated rare exon CNVs using 14 coverage- and quality-derived features. On 492 real WES samples and NA12878 gold-standard data, ML-ExonCNV outperformed GATK-gCNV, ExomeDepth, and CNVkit. The tool detects large segmental, mosaic, and breakpoint-spanning CNVs at exon resolution, expanding clinically relevant CNV discovery from exomes. 3

Long reads plus T2T-CHM13 substantially expand diagnostic-grade structural variant detection

This multiplatform study compared SV detection using short-read sequencing, long-read sequencing, and optical genome mapping on hg38 versus T2T-CHM13. Most SVs called by short reads were confirmed by long reads, but long reads identified about twice as many SVs per genome. Using T2T-CHM13 increased deletion calls by 20% and reduced insertion calls by 20% compared with GRCh38, especially for long-read data. About 80% of SVs detected by short and long reads were under 0.5 kb, largely invisible to optical genome mapping. 4