CD-GPT As a Biological Foundation Model Bridging the Gap between Molecular Sequences Through Central Dogma

• DOI: 10.1101/2024.06.24.600337
• Publication Date: 2024-06-29T00:40:32Z
• Authors: Xiao Zhu; Chenchen Qin; Fang Wang; Fan Yang; Bing He; Yu Zhao; Jianhua Yao
• Keywords: Foundation (evidence)

Abstract The central dogma serves as a fundamental framework for understanding the flow and expression of genetic information within living organisms, facilitating connection diverse biological sequences across molecule types. In this study, we present CD-GPT (Central Dogma Generative Pretrained Transformer), generative foundation model comprising 1 billion parameters, aiming to capture intricate system-wide molecular interactions in systems. We introduce concept unified representational space employ shared, multi-molecule vocabulary effectively represent narrow their distance embedding space. Through extensive pretraining on comprehensive full level data, exhibits exceptional performance wide range predictive downstream tasks, encompassing mono-molecular multi-molecular analyses. Notably, excels tasks such genomic element detection, protein property prediction, RNA-protein interaction identification also like de novo generation reverse translation. versatility opens up promising avenues advanced multi-omics analysis.