Dataset from:An Evaluation of Large Language Models in Bioinformatics Research

This repository contains the data and code to reproduce the results of our paper: An Evaluation of Large Language Models in Bioinformatics Research Authors: Hengchuang Yin; Lun Hu Abstract: Large language models, such as the GPT series, have revolutionized natural language processing by demonstrating strong capabilities in text generation and reasoning. However, their potential in the field of bioinformatics, characterized by complex biological data and specialized knowledge, has not been fully evaluated. In this study, we systematically assess the performance of multiple advanced and widely used LLMs on six diverse bioinformatics tasks: drug-drug interaction prediction, antimicrobial and anticancer peptide identification, molecular optimization, gene and protein named entity recognition, single-cell type annotation, and bioinformatics problem solving.Our experimental results demonstrate that, with appropriate prompt design and limited task-specific fine-tuning, general-purpose LLMs can achieve competitive or even superior performance compared to traditional models that require extensive computational resources and technical design across various tasks. Our analysis further uncovers the current limitations of LLMs in handling structurally complex and knowledge-intensive bioinformatics problems. Overall, this study demonstrates the broad prospects of LLMs in bioinformatics while emphasizing their limitations, providing valuable insights for future research at the intersection of LLMs and bioinformatics. Section_A_ddi: ddinter_positive_samples.csv: Positive drug–drug interaction (DDI) pairs curated from the DDInter database. ddinter_negative_samples.csv: Negative drug–drug interaction (DDI) pairs (no known interactions), used for supervised classification. drug_description_embeddings_all-mpnet-base-v2.npy: Drug description embeddings generated using the all-mpnet-base-v2 model. drug_description_embeddings_bge-large-en-v1.5.npy: Drug description embeddings generated using the bge-large-en-v1.5 model. drug_description_embeddings_e5-small-v2.npy: Drug description embeddings generated using the e5-small-v2 model. drug_description_embeddings_gtr-t5-large.npy: Drug description embeddings generated using the gtr-t5-large model. drug_description_embeddings_text_embedding_3_large.npy: Drug description embeddings generated using OpenAI's text-embedding-3-large model. drug_description_embeddings_text_embedding_3_small.npy: Drug description embeddings generated using OpenAI's text-embedding-3-small model. drug_description_embeddings_text_embedding_ada_002.npy: Drug description embeddings generated using OpenAI's text-embedding-ada-002 model. We hereby confirm that the dataset associated with the research described in this work is made available to the public under the Creative Commons Zero (CC0) license. Contact If you have any questions, please don't hesitate to ask me: [e-mail] or [e-mail]