WHAT IS THE IMPACT OF COMPOUND HETEROZYGOUS EVENTS INVOLVING DELETIONS AND SEQUENCE-LEVEL VARIANTS IN AUTISM?

The majority of human genes maintain normal biological function when they become haploid due to a genomic deletion. However, pathogenicity may still arise when the remaining allele is affected by additional functional variation. Here, we describe analytical strategies for examining a specific type of compound heterozygosity, namely the co-occurrence of a deletion and a sequence-level variant affecting the other allele, hereafter referred to as deletion compound heterozygosity (DelCH). We report preliminary results in using these strategies to assess DelCH in Autism Spectrum Disorder (ASD).

We analyzed whole-genome sequencing data from MSSNG, Simons Simplex Collection, and SPARK cohorts (collectively 11,636 autistic individuals and 22,962 family members).

We developed multiple analytical strategies to examine rare (event rate < 1%) DelCH:

• 1)
  The burden analysis uses conditional logistic regression for group-level comparisons of DelCH rates between a) probands and their deletion-transmitting parents, with inherited deletion as a random effect variable, or b) probands and their family members, with family ID as a random effect variable;
• 2)
  The transmission disequilibrium test (TDT) compares the rates with which deletion-non-transmitting parents transmit sequence-level variants within genes affected by deletions to their autistic offspring. Association is indicated by transmission of non-synonymous variants at a rate higher than predicted by chance. This approach was repeated in unaffected siblings as an additional control analysis.

Each strategy has different strengths and weaknesses. The first burden analysis (1a) achieves perfect matching of deleted sequence and unambiguous phasing of variants but is restricted to proband-parent pairs. The second burden analysis (1b) benefits from a larger sample size but cannot distinguish between de novo and inherited variation. In addition, unambiguous phasing is possible only for SNVs within deletion boundaries. In contrast, while the TDT (2) can include SNVs outside deletion boundaries, thereby increasing statistical power, de novo events are not analyzed.

Our preliminary findings show variability in results as a function of the analytical strategy. Findings from the burden analysis suggest a modest enrichment of DelCH in ASD which was inversely proportional to the variant frequency thresholds applied.

Given that the mechanism consists of two rare events at the same locus, on the population level the role of DelCH in ASD etiology is likely modest, requiring large samples sizes for sufficient statistical power. In addition to this “lightning striking twice”, data preparation is demanding, as every subject has unique deletion regions in which sequence-level variants on the other allele are tallied. Variant selection metrics include allele frequency thresholds, the threshold for predicted autosomal dominant deleteriousness, and variant inclusion rules with respect to the deletion boundaries. Our findings show that altering any of these metrics affects the results, warranting careful consideration.

Understanding the potential role of DelCH may contribute to a more comprehensive library of ASD-associated genetic variation. DelCHs could explain why individual probands with ASD present with a deletion inherited from a non-affected parent. Lastly, examining DelCH in genetic analyses may help identify recessive ASD genes, which otherwise escape detection.