We’ve updated our EVEscape paper with more SARS2 retrospective evaluation and potential use-cases! Collaborators @sarahgurev @NotinPascal @nooryoussef03 @_nathanrollins @sandercbio @yaringal (@OATML_Oxford), @deboramarks, 🧵1/4.

RT @AprioriBio: We recently spoke with senior scientist Nicole Thadani, PhD, about her interest in immunology and how her work is motivated….

RT @roseorenbuch: Announcing popEVE - a deep generative model of the human proteome that reveal over a hundred novel genes involved in rare….

Thrilled to share EVEscape in its final form and grateful to all coauthors, collaborators and reviewers!

RT @H__Spinner: Forecasting viral evolution with pre-pandemic sequences alone !!! Really cool work from @NthadaniN @sarahgurev @NotinPascal….

We provide scores for emerging strains and potential future mutations with monthly updates at Code to compute scores is at We’re now diving into applying EVEscape for crucial pandemic preparedness challenges! 4/4.

Going forward: EVEscape shows promise for identifying SARS2 strains with high escape potential as they emerge. We explore extensions - predicting effects of indels and glycosylation, and modeling escape for viruses with future pandemic potential. 3/4

Looking back: EVEscape, using data available before 2020, predicts pandemic mutations on par with experimental approaches that require host sera samples, outperforms computational approaches that rely on pandemic sequencing, and excels at predicting high-frequency mutations. 2/4

RT @deboramarks: Primer on Medical and Population Genetics | Broad Institute ⁦ - Nikki Thadani and Sarah Gurev from my lab - suggesting a….

RT @the_tessallator: Introducing scPerturb, a collection of 44 single cell perturbation datasets with unified annotations available at http….

Nice compbio tools list from one of my labmates - check it out!.

We will flag new strains with high escape potential and mutations on top of these strains at Looking forward to expanding our approach to take advantage of the ever-increasing wealth of pandemic data! 10/10.

EVEscape scores for all mutants to SARS-CoV-2 Spike and strain-level scores for all PANGO lineages can be found in the supplement. Code to compute scores is at 9/10.

We extend this approach to evaluate the escape potential of strains; Omicron variants like BA.4 stand out as having particularly high escape potential, even when compared to combinations of mutations already seen in VOCs. 8/10

A third of mutations seen frequently in historical VOCs are in the top 10% of Spike EVEscape predictions. The top EVEscape-predicted RBD site is E484, which has confirmed escape mutations in the VOCs including Omicron (resulting in bamlanivimab escape). 7/10.

Promisingly, model predictions align with trends in the SARS-CoV-2 pandemic - mutations with higher EVEscape scores are more likely to be observed, particularly as the pandemic progresses. 6/10

We evaluate EVEscape against the tour de force of DMS studies that measure RBD antibody and sera binding (@jbloom_lab, Sunney Xie lab) and find that of the top 10% of EVEscape site predictions, 50% are observed sites of sera escape. 5/10

We developed EVEscape, quantifying immune escape propensity using three modules: deep learning model of broader viral evolution + protein 3D structure + residue dissimilarity. The model does NOT use pandemic viral sequencing or known Ab binding. 4/10

Although sequence models of broader evolution are surprisingly useful for clinically relevant human variant effect prediction (, they don’t necessarily capture information about which mutations are likely to disrupt binding of pre-existing antibodies. 3/10.

As SARS-CoV-2 has demonstrated, virus evolution can readily outpace the development of vaccines and therapeutic antibodies. We made predictions of viral evolution exploiting the power of evolutionary and structural information available at the start of the pandemic. 2/10.