ELOVL2ELOVL fatty acid elongase 2

Krishnan and colleagues generated probability scores genome-wide by using a machine learning approach on a human brain-specific gene network. The method was first presented in Nat Neurosci 19, 1454-1462 (2016), and scores for more than 25,000 RefSeq genes can be accessed in column G of supplementary table 3 (see: http://www.nature.com/neuro/journal/v19/n11/extref/nn.4353-S5.xlsx). A searchable browser, with the ability to view networks of associated ASD risk genes, can be found at asd.princeton.edu.

The Exome Aggregation Consortium (ExAC) is a summary database of 60,706 exomes that has been widely used to estimate 'constraint' on mutation for individual genes. It was introduced by Lek et al. Nature 536, 285-291 (2016), and the ExAC browser can be found at exac.broadinstitute.org. The pLI score was developed as measure of intolerance to loss-of- function mutation. A pLI > 0.9 is generally viewed as highly constrained, and thus any loss-of- function mutations in autism in such a gene would be more likely to confer risk. For a full list of pLI scores see: ftp://ftp.broadinstitute.org/pub/ExAC_release/release0.3.1/functional_gene_constraint/fordist_cle aned_exac_nonTCGA_z_pli_rec_null_data.txt

The TADA score ('Transmission and De novo Association') was introduced by He et al. PLoS Genet 9(8):e1003671 (2013), and is a statistic that integrates evidence from both de novo and transmitted mutations. It forms the basis for the claim of 65 individual genes being strongly associated with autism risk at a false discovery rate of 0.1 (Sanders et al. Neuron 87, 1215-1233 (2015)). The calculated TADA score for 18,665 RefSeq genes can be found in column P of Supplementary Table 6 in the Sanders et al. paper (the column headed 'tadaFdrAscSscExomeSscAgpSmallDel'), which represents a combined analysis of exome data and small de novo deletions (see www.cell.com/cms/attachment/2038545319/2052606711/mmc7.xlsx).

The DAMAGES score (disease-associated mutation analysis using gene expression signatures), or D score, was developed to combine evidence from de novo loss-of- function mutation with evidence from cell-type- specific gene expression in the mouse brain (specifically translational profiles of 24 specific mouse CNS cell types isolated from 6 different brain regions). Genes with positive D scores are more likely to be associated with autism risk, with higher-confidence genes having higher D scores. This statistic was first presented by Zhang & Shen (Hum Mutat 38, 204- 215 (2017), and D scores for more than 20,000 RefSeq genes can be found in column M in supplementary table 2 from that paper.