Supplementary Information for Predicting the clinical impact of human mutation with deep neural networks

By Laksshman Sundaram, Hong Gao, S. Padigepati, J. McRae, Yanjun Li, J. Kosmicki, Nondas Fritzilas, Jörg Hakenberg, Anindita Dutta, J. Shon and 4 more. 2018

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

1. Jimin Pei
2. Max Schubach
3. Lisa N. Kinch
4. Nick V. Grishin
5. Joshua Meier
6. Roshan Rao
7. Jun Cheng
8. Tom Sercu

Citations in the corpus

• MSA Transformer (2021, neutral): “…fit to MSAs are widely used in computational biology including in applications such as fitness landscape prediction (Riesselman et al., 2018), pathogenicity prediction (Sundaram et al., 2018; Frazer et al., 2020), remote homology detection (Hou et al., 2018), and protein design (Russ et al., 2020).”
• Language models enable zero-shot prediction of the effects of mutations on protein function (2021, methodology): “Models have been designed specifically for proteins, using feature engineering such as Envision [38] and PolyPhen-2 [39], ensemble methods such as Revel [40], MPC [41], CADD [42], and M-CAP [43], language models such as UniRep [21, 16] and ESM [12], and other representation learning approaches [44, ”
• LIST-S2: taxonomy based sorting of deleterious missense mutations across species (2020, methodology): “0a (17) (namely: SIFT (1), SIFT4G (18), PROVEAN (2), phyloP (3) X 3, SiPhy (5), phastCons (6) X 3, GERP++ (4), Eigen (10), CADD (9), DANN (11), W158 Nucleic Acids Research, 2020, Vol. 48, Web Server issue PolyPhen-2 (8), FATHMM-MKL (19), PrimateAI (20), MutationTast (21), MutationAssessor (22), MPC ”
• Mutation severity spectrum of rare alleles in the human genome is predictive of disease type (2019, methodology): “Natural variants with clinical impact tend to be rare in the human population [48].”
• Accurate proteome-wide missense variant effect prediction with AlphaMissense (2023, methodology): “For the training set, we assign benign labels to variants frequently observed in the human and primate populations, and pathogenic labels to variants absent from human and primate populations, as is done in PrimateAI ( 12 ) (Fig.”
• Ensembl 2022 (2021, methodology): “This year we created extensions to integrate variant pathogenicity predictions from ClinPred (31) and PrimateAI (32).”
• Artificial intelligence in clinical and genomic diagnostics (2019, methodology): “One approach, PrimateAI, which used CNNs trained on variants of known pathogenicity with data augmentation using cross-species information, was shown to outperform prior methods when trained directly upon sequence alignments [59].”
• CADD v1.7: using protein language models, regulatory CNNs and other nucleotide-level scores to improve genome-wide variant predictions (2024, background): “Further, the broader concept of using species differences and long-standing variation has been applied to develop improved models of missense effects ( 30 ,31 ).”
• dbNSFP v4: a comprehensive database of transcript-specific functional predictions and annotations for human nonsynonymous and splice-site SNVs (2020, background): “…Eigen and Eigen PC [19], FATHMM-XF [20], GenoCanyon [21], LINSIGHT [22], LIST-S2 [23], M-CAP [24], MPC [25], MutPred [26], MVP [27], PrimateAI [28], REVEL [29], SIFT4G [30]), one score for loss of function prediction (ALoFT [31]), and three conservation scores (phyloP17-way_primate [32],…”