Autism spectrum disorder (ASD) is caused by combined genetic and environmental factors. Genetic heritability in ASD is estimated as 60-90%, and genetic investigations have revealed many monogenic factors. We analyzed 405 patients with ASD using family-based exome sequencing to detect disease-causing single-nucleotide variants (SNVs), small insertions and deletions (indels), and copy number variations (CNVs) for molecular diagnoses. All candidate variants were validated by Sanger sequencing or quantitative polymerase chain reaction and were evaluated using the American College of Medical Genetics and Genomics/Association for Molecular Pathology guidelines for molecular diagnosis. We identified 55 disease-causing SNVs/indels in 53 affected individuals and 13 disease-causing CNVs in 13 affected individuals, achieving a molecular diagnosis in 66 of 405 affected individuals (16.3%). Among the 55 disease-causing SNVs/indels, 51 occurred de novo, 2 were compound heterozygous (in one patient), and 2 were X-linked hemizygous variants inherited from unaffected mothers. The molecular diagnosis rate in females was significantly higher than that in males. We analyzed affected sibling cases of 24 quads and 2 quintets, but only one pair of siblings shared an identical pathogenic variant. Notably, there was a higher molecular diagnostic rate in simplex cases than in multiplex families. Our simulation indicated that the diagnostic yield is increasing by 0.63% (range 0-2.5%) per year. Based on our simple simulation, diagnostic yield is improving over time. Thus, periodical reevaluation of ES data should be strongly encouraged in undiagnosed ASD patients.
Many algorithms to detect copy number variations (CNVs) using exome sequencing (ES) data have been reported and evaluated on their sensitivity and specificity, reproducibility, and precision. However, operational optimization of such algorithms for a better performance has not been fully addressed. ES of 1199 samples including 763 patients with different disease profiles was performed. ES data were analyzed to detect CNVs by both the eXome Hidden Markov Model (XHMM) and modified Nord's method. To efficiently detect rare CNVs, we aimed to decrease sequencing biases by analyzing, at the same time, the data of all unrelated samples sequenced in the same flow cell as a batch, and to eliminate sex effects of X-linked CNVs by analyzing female and male sequences separately. We also applied several filtering steps for more efficient CNV selection. The average number of CNVs detected in one sample was <5. This optimization together with targeted CNV analysis by Nord's method identified pathogenic/likely pathogenic CNVs in 34 patients (4.5%, 34/763). In particular, among 142 patients with epilepsy, the current protocol detected clinically relevant CNVs in 19 (13.4%) patients, whereas the previous protocol identified them in only 14 (9.9%) patients. Thus, this batch-based XHMM analysis efficiently selected rare pathogenic CNVs in genetic diseases.