LEGENDS OF FIGURES
Figure 1: Workflow of the exome sequencing analysis. Variants were filtered according to quality (Phred score >20, read depth >10, variant allele frequency >35%); coding and noncoding variants were separately analyzed; frequency (GnomAD, ABraOM, 1K Genomes); the effect on coding variants (frameshift, stop loss/gain, missense, splice site, nonsense); for missense variants, the prediction of pathogenicity in at least 5 out of 6 algorithms; HPO annotation (hepatoblastoma, abnormalities of the liver, and cancer). The filtered variants were visually examined using Integrative Genomics Viewer (IGV) software (http//www.broadinstitute.org/igv) to further filter out possible strand bias and homopolymeric region artifacts. All the filtered variants mapped to cancer predisposition genes were annotated using the ACMG guidelines. 1- Intronic variants, 3’UTR, 5’UTR; 2- MAF: GnomAD, ABraOM, 1K Genomes; 3- Frameshift, stop loss, stop gain, missense, splice site, nonsense variants; 4- Missense variants with dbNSFP Functional Prediction of pathogenicity in at least 5 out of 6 algorithms; 5- Terms used for Varelect and HPO annotation: Hepatoblastoma, abnormalities of the liver, and cancer. CNV - copy number variation, ROH - region of homozygosity, MAF – maximum allele frequency, CPG – cancer predisposition gene, VUS – variant of uncertain significance, HPO - Human Phenotype Ontology, HB - hepatoblastoma.