Reproduce “Survey and Improvement Strategies for Gene Prioritization with Large Language Models” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Matthew Neeley, Guantong Qi, Guanchu Wang, Ruixiang Tang, Dongxue Mao, Chaozhong Liu, Sasidhar Pasupuleti, Bo Yuan and 4 more
This study evaluates the effectiveness of Large Language Models (LLMs) in prioritizing disease-causing genes for rare genetic diseases by analyzing patient phenotypes and candidate gene lists. Researchers identified and addressed inherent biases in LLM performance by developing a 'divide-and-conquer' strategy, leading to improved gene prioritization accuracy, particularly for larger gene sets.
Key insights
- Large Language Models exhibit biases in gene prioritization, notably a 'literature bias' towards well-studied genes and a 'positional bias' where genes later in the input are ranked lower.
- A 'divide-and-conquer' strategy, breaking down large gene lists into smaller subsets for LLM processing, effectively mitigates these biases and improves performance.
- The inherent 'solvability' of patient cases, often correlated with the specificity of phenotypic descriptions, influences LLM performance in gene prioritization.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.