De novo mutations in autism spectrum disorders and an empirical test of the neutral DNA model

Abstract The genetic basis of autism spectrum disorders (ASD) remains better understood and might concern only a small fraction of the genome if the neutral theory were true. We here analyzed published de novo single nucleotide variations in ASD and controls to better understand de novo mutations (DNMs) and to empirically test the neutral model. By comparing DNMs in ASD and control subjects with single nucleotide polymorphisms (SNPs) found in the 1000 genomes project, we found that DNMs in normal subjects occurred at SNP sites at least 3.45 fold more frequent than expected from the neutral theory, whereas DNMs in ASD were less frequent (2.81 fold), especially so for common SNPs with minor allele frequency >0.01 (0.97 fold in ASD vs 1.81 fold in normal). We also analyzed the nucleotide compositions of DNMs and the parity (1:1 ratio) of pyrimidines and purines (Chargaff Parity Rule 1 and 2). We found that DNMs in ASD showed overall lower AT content than that in controls (P<0.05). Parity violations and AT bias in DNMs occurred at expected frequency based on chance (~0.05) in both ASD and controls. However, no individual was found to have both parity violation and AT bias among 290 normal and 429 ASD subjects studied. These results show enrichment of DNMs in SNP sites in normal subjects and less so in ASD, which is not expected from the neutral model, and indicate that DNMs are on average more deleterious in ASD than in controls.