Disease variant prediction with deep generative models of evolutionary data
By J. Frazer, Pascal Notin, Mafalda Dias, Aidan N. Gomez, Joseph K. Min, Kelly P. Brock, Yarin Gal, Debora Marks. Nature, 2021
734 citations recorded by the source corpus.
Quantifying the pathogenicity of protein variants in human disease-related genes would have a marked effect on clinical decisions, yet the overwhelming majority (over 98%) of these variants still have unknown consequences1–3. In principle, computational methods could support the large-scale interpretation of genetic variants. However, state-of-the-art methods4–10 have relied on training machine learning models on known disease labels. As these labels are sparse, biased and of variable quality, the resulting models have been considered insufficiently reliable11. Here we propose an approach that leverages deep generative models to predict variant pathogenicity without relying on labels. By modelling the distribution of sequence variation across organisms, we implicitly capture constraints on the protein sequences that maintain fitness. Our model EVE (evolutionary model of variant effect) not only outperforms computational approaches that rely on labelled data but also performs on par with, if not better than, predictions from high-throughput experiments, which are increasingly used as evidence for variant classification12–16. We predict the pathogenicity of more than 36 million variants across 3,219 disease genes and provide evidence for the classification of more than 256,000 variants of unknown significance. Our work suggests that models of evolutionary information can provide valuable independent evidence for variant interpretation that will be widely useful in research and clinical settings. A new computational method, EVE, classifies human genetic variants in disease genes using deep generative models trained solely on evolutionary sequences.
Subfields: Anthropology, Visual Arts and Performing Arts
View Paper: https://www.semanticscholar.org/paper/0537c62e32299b8e9b4e3cac44d82ffa123cbbe1
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
- Erik Nijkamp
- Jeffrey A. Ruffolo
- Kevin Kaichuang Yang
- Sergey Levine
- Pushmeet Kohli
- Fajie Yuan
- Ali Madani
- M. Amode
Citations in the corpus
- Genome-wide prediction of disease variants with a deep protein language model (2022, methodology): “Over the past decades, computational variant functionalization tools have dramatically improved in accuracy and usability [4].”
- Exploring evolution-based & -free protein language models as protein function predictors (2022, background): “Our work points out both strengths and weakness of three SOTA PLMs, which are potentially useful to many downstream biological tasks, including but not limited to protein engineering [6], cancer early detection [14], and drug discovery [21].”
- Ensembl 2023 (2022, methodology): “We have continued to extend the available analysis options and have this year integrated predictions from EVE (38), which employs an evolutionary model of variant effect, and CAPICE (39), a machine-learning based method for prioritising pathogenic variants.”
- ProGen2: Exploring the Boundaries of Protein Language Models (2022, methodology): “Extending this approach with a Bayesian VAE, EVE (Frazer et al., 2021) achieved further improvements to variant effect prediction.”
- SaProt: Protein Language Modeling with Structure-aware Vocabulary (2024, methodology): “In our analysis, we harness the data sourced from EVE (Frazer et al., 2021), additionally filtering out proteins with length greater than 1024 or absent from the AlphaFoldDB.”
- Accurate proteome-wide missense variant effect prediction with AlphaMissense (2023, methodology): “AlphaMissense pathogenicity correlates with MAVE data measuring the fitness of auxotrophic yeast strains expressing human GCK variants in the presence of glucose ( 45 ) (Spearman correlation: 0.53), outperforming ESM1v, EVE, and ESM1b (Spearman correlation: 0.49, 0.48, and 0.45, respectively; fig.”
- Electroencephalographic insights into variant function and clinical outcomes in SCN2A encephalopathy (2023, methodology): “24 Second, an in-silico pathogenicity prediction score was included; this score was recently developed specifically to assess the pathogenicity of variants in voltage-gated sodium channels.”