FUSE: Revolutionizing the analysis of gene mutations

How would you summarize your study for a lay audience? 

Our study introduces a new tool called FUSE (Functional Substitution Estimation) that helps scientists better understand how changes in genes affect proteins. Genetic variants can alter how a protein works, potentially leading to diseases. New CRISPR-based experiments can help us understand the impact of genetic variants by installing these changes into the DNA of cells. FUSE combines data from many of these experiments to more accurately predict the impact of specific genetic changes, even for mutations that haven’t been tested yet. This advancement can provide better evidence to understand the impacts of genetic variants, which can ultimately help doctors identify harmful mutations more effectively, leading to improved patient care and personalized treatments.

What knowledge gap does your study help to fill? 

In this study, we aimed to improve the accuracy of interpreting how genetic mutations affect protein function, which is crucial for understanding disease risk. High-throughput functional screening assays, like deep mutational scanning, produce large amounts of data on how mutations impact proteins. However, individual measurements from these assays can be noisy due to experimental variability, making it challenging to precisely estimate the effect of each variant. 

What drove you to pursue research in this area? 

In recent years, there has been a significant increase in studies aiming to experimentally measure the impact of genetic variants across the human genome. These efforts have produced hundreds of thousands of screening results from high-throughput functional assays. However, each individual measurement can be affected by statistical noise and experimental variability, which may limit the accuracy of the estimates when considered in isolation. 

Motivated by this challenge, our research teams recognized an opportunity to enhance the accuracy of these variant impact estimates by collectively analyzing the vast amount of available data. By integrating results from numerous studies, we aimed to reduce the noise inherent in individual experiments and improve the reliability of each estimate. 

Our two labs combined computational and experimental expertise to tackle complex problems in genomics. This synergy allowed us to develop a new approach that leverages both computational methods and experimental data to refine findings. Our goal was to create a tool that not only advances our understanding of genetic variants but also provides a valuable resource for the scientific community, ultimately contributing to improved patient care and personalized medicine. 

What methods or approaches did you use? 

To address this, we developed FUSE. We collected and integrated data from over 100 functional experimental screening datasets covering numerous genes. By analyzing this extensive dataset collectively, FUSE reduces noise and improves the accuracy of functional impact estimates for each variant. We also created a new amino acid substitution matrix called FUNSUM, derived from high-quality and de-noised data, which helps adjust for expected functional impacts at the residue level. 

What did you find? 

Our findings demonstrated that FUSE significantly enhances the reliability of functional estimates, improves the classification of pathogenic and benign variants in clinical databases like ClinVar, and better predicts disease risk in patients with rare variants, as shown using data from the UK Biobank. 

What are the implications? 

Our work has significant clinical implications for patient care and precision medicine. By providing more accurate assessments of genetic variant impacts, FUSE can help clinicians and genetic counselors better distinguish between harmful and benign mutations, particularly those currently classified as “variants of uncertain significance” (VUS). This improved interpretation can lead to more accurate diagnoses, personalized risk assessments, and informed decision-making regarding prevention strategies and treatment options for patients. 

Additionally, by inputting the effects of unscreened variants where some screening has already occurred nearby, FUSE addresses gaps caused by limitations in experimental assays, expanding the range of variants that can be evaluated. This means we can provide reliable functional impact estimates even for mutations that haven’t been directly tested in the lab. 

What are the next steps? 

The next steps include applying FUSE to a wider array of functional screening approaches, such as emerging platforms like base and prime editing. We aim to further measure the strength of clinical support provided by our imputed functional scores which have not been originally screened. By collaborating with the broader scientific and medical communities, we hope to integrate FUSE into existing tools and databases, ultimately contributing to improved patient outcomes through enhanced precision medicine.