The results confirm the feasibility of predicting fitness effect of a selected base editor-sgRNA combination using sequence as the primary feature, opening the way for future studies to incorporate additional biological features to enhance model interpretability and transferability.
Abstract
Pooled CRISPR base-editing screens enable the functional assessment of large numbers of genetic variants at scale. Predicting whether a selected sgRNA-Base Editor combination will produce a robust fitness effect could support sgRNA prioritisation and reveal features that are transferrable across screening experiments. However, this remains challenging due to the number of molecular and biological events taking place in a screening experiment. Here we investigated whether sgRNA fitness effect can be predicted by Machine Learning from a set of biological and sequence-specific features using a published Base Editing screening dataset. Decision Tree based models showed AUC-ROC ranging from 87 to 94% and AU-PR ranging from 53 to 75% using a gene-held out validation. Feature interpretability analysis revealed that sgRNA sequence, in particular PAM-proximal nucleotides, were the most predictive features. Gene essentiality and Predicted Variant Effect also carried predictive weight albeit to a more marginal extent. Analysis of matched Cas9 and base-editing data further identified gene essentiality as the feature most strongly associated with the different fitness score among base editing and Cas9. Our results confirm the feasibility of predicting fitness effect of a selected base editor-sgRNA combination using sequence as the primary feature, opening the way for future studies to incorporate additional biological features to enhance model interpretability and transferability.
Built with VRH, VenusREM2 is the first to rank highest in all function, taxon, MSA-depth, and mutation-depth categories, with a ProteinGym Average Spearman of 0.556, 0.038 above the prior best.
Yang Tan, Qi-Jia Tian, Gang-Yu Sun et al.· arXiv.org· 1 citation
This work shows that redundancy among partially overlapping deep mutational scanning experiments, often treated as noise, yields a consensus landscape of mutational constraint, and introduces "variant soundness", an overlap-aware rank-fusion score that uses cross-experiment consistency to derive an assay-agnostic order...
Riccardo Arnese, G. Gambardella· Molecular Systems Biology· 0 citations
Machine learning-based annotation methods are increasingly used to assess the pathogenicity of genetic variants, but their performance at prioritizing variants for gene-level association testing remains poorly characterized. Here, to better understand and optimize for this use case, we assess variant annotations from...
Matthew Aguirre, Flaviyan Jerome Irudayanathan, M. Crow et al.· BMC Genomics· 0 citations
Deep learning sequence-to-function (S2F) models learn to predict functional readouts, such as gene expression, from genetic sequence. However, these models are trained predominantly on the reference genome and struggle to accurately predict the effect of genetic variation on gene expression. Massively parallel reporter...
Lena Wang, Jake Galvin, Eduarda Vaz et al.· bioRxiv· 0 citations
Most common variants occur in non-coding genomic regions. Numerous expression quantitative trait loci (eQTLs) can influence nearby gene expression. Across diverse traits and diseases, many genome-wide association study (GWAS) loci have been shown to colocalize with eQTLs, supporting gene-regulatory causal mechanisms. H...
Andreas W. Schmidt, Jonas Rosendahl, Helmut Laumen et al.· bioRxiv· 0 citations
Related blog posts
MIT News · Artificial Intelligence· news.mit.eduNov 18, 2026
What's new in research at Microsoft? Host Anson Ho explores the latest, including AI-powered universal translation, a breakthrough in a 31-year-old mathematics challenge, and advances making AI coding agents faster and more efficient. The post Microsoft Research Focus – Edition #98 appeared first on Microsoft Research.
MIT News · Artificial Intelligence· news.mit.eduNov 12, 2026
MIT scientists developed a training method to generate models that simulate the effects of human activities on climate much more quickly than other approaches.
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.