Systematic analysis of homozygous autosomal copy number losses in exomes improves diagnostic yield and uncovers ultra-rare recessive disorders

Systematic analysis of copy number variants (CNVs) in large datasets is challenging and there are limited studies of homozygous copy number losses in rare disease exome datasets. Here we leveraged the genomic uniqueness and relative under-representation of the Indian population in the current public genomic databases and identified 42,386 possible homozygous losses (median count 20 per individual, range 0 - 55; median size 2.95 kb, range 99 bp - 4.76 Mb) in a heterogeneous cohort of 2,021 individuals with suspected Mendelian disorders, who had undergone exome sequencing using 12 different capture kits in a resource-limited setting. Employing a genomic position loss-count based approach, we filtered 1,224 rare homozygous loss calls in 718 individuals (median count 1 per individual, range 0 - 22; median size 3.49 kb, range 121 bp - 4.76 Mb) for further analysis, thus significantly reducing the analysis burden. Clinical correlation and validation of these rare calls enabled 10 new diagnoses in 240 unsolved individuals with at least one filtered rare homozygous loss call. This, led to nearly two-fold increase in diagnosis owing to homozygous deletions in our cohort. Further analysis of the data and identification of additional affected individuals through collaboration led to identification of biallelic <i>FILIP1</i> and <i>FAM177A1</i> variants as causes of a syndromic arthrogryposis and a neuromuscular disorder respectively. Both these conditions have been recently proven as ultra-rare recessive disorders, thus validating our approach. We also show that biallelic loss-of-function <i>TFCP2L1</i> variants cause chronic kidney disease and <i>VPS36</i> variants cause a severe recessive neurodevelopmental disorder characterised by microcephaly, motor delay, agenesis of the corpus callosum, cerebellar atrophy, seizures, hypotonia, spasticity and early death. Overall, these results demonstrate a scalable approach to screen homozygous losses for improving diagnostic yield and discovering disease-genes in large exome cohorts.