Trio-whole-exome sequencing analysis

The 37 Italian patient-parents trios and German patients underwent whole-exome sequencing (WES). Exome library capture and enrichment was performed with BGI kit (BGI Genomics, n = 104) or the NimbleGen SeqCap EZ (Roche, n = 7) and sequenced as paired-end reads with an Illumina system. As previously described,²³ sequencing reads were processed to obtain annotated variant calls and retain for downstream analyses only variants located in coding exons or canonical splice-site junctions. Guidelines for sequence variant interpretation of the American College of Medical Genetics (ACMG)²⁴ were followed for clinical classification of variants in established epilepsy genes by using the online platform Varsome (https://varsome.com/).

We calculated the mean rate of de novo variants (DNVs) in the main variant classes (missense, nonsense, splicing affecting, frame-shift, in-frame indels and synonymous single nucleotide variants) and used for comparison data from a large control cohort of 1911 trios previously published.²⁵ Although we did not formally account for differences in variant discovery workflows,²⁵ they have similarities in both enrichment methods and bioinformatic pipelines.

To select candidate FE genes, we used the following criteria: (i) the DNV has allele count ≤1 across control samples in the genome aggregation database (gnomAD) data set v2.1.1 including counts on 60146 chromosomes; (ii) the DNV has CADD Phred score (v1.4)  ≥ 15; (iii) the gene is expressed in brain according to the Human Protein Atlas and Human Allen Brain Atlas databases; (iv) the encoded protein is important for brain development and function based on NCBI Entrez Gene Summary and UniprotKB databases as well as bibliographic research; and (v) the gene is intolerant to variation according to gnomAD v2.1.1 measures for genetic constraints (missense Z score > 3 and/or pLI > 0.9) or already associated with epileptic seizures in humans.