In recent years, significant progress has been made in the field of protein function prediction with the development of various machine-learning approaches. However, most existing methods formulate the task as a multi-classification problem, i.e. assigning predefined labels to proteins. In this work, we propose a novel approach, Prot2Text, which predicts a protein's function in a free text style, moving beyond the conventional binary or categorical classifications. By combining Graph Neural Networks(GNNs) and Large Language Models(LLMs), in an encoder-decoder framework, our model effectively integrates diverse data types including protein sequence, structure, and textual annotation and description. This multimodal approach allows for a holistic representation of proteins' functions, enabling the generation of detailed and accurate functional descriptions. To evaluate our model, we extracted a multimodal protein dataset from SwissProt, and demonstrate empirically the effectiveness of Prot2Text. These results highlight the transformative impact of multimodal models, specifically the fusion of GNNs and LLMs, empowering researchers with powerful tools for more accurate function prediction of existing as well as first-to-see proteins.

BACKGROUND: Genome layout and gene regulation appear to be interdependent. Understanding this interdependence is key to exploring the dynamic nature of chromosome conformation and to engineering functional genomes. Evidence for non-random genome layout, defined as the relative positioning of either co-functional or co-regulated genes, stems from two main approaches. Firstly, the analysis of contiguous genome segments across species, has highlighted the conservation of gene arrangement (synteny) along chromosomal regions. Secondly, the study of long-range interactions along a chromosome has emphasised regularities in the positioning of microbial genes that are co-regulated, co-expressed or evolutionarily correlated. While one-dimensional pattern analysis is a mature field, it is often powerless on biological datasets which tend to be incomplete, and partly incorrect. Moreover, there is a lack of comprehensive, user-friendly tools to systematically analyse, visualise, integrate and exploit regularities along genomes. RESULTS: Here we present the Genome REgulatory and Architecture Tools SCAN (GREAT:SCAN) software for the systematic study of the interplay between genome layout and gene expression regulation. GREAT: SCAN is a collection of related and interconnected applications currently able to perform systematic analyses of genome regularities as well as to improve transcription factor binding sites (TFBS) and gene regulatory network predictions based on gene positional information. CONCLUSIONS: We demonstrate the capabilities of these tools by studying on one hand the regular patterns of genome layout in the major regulons of the bacterium Escherichia coli. On the other hand, we demonstrate the capabilities to improve TFBS prediction in microbes. Finally, we highlight, by visualisation of multivariate techniques, the interplay between position and sequence information for effective transcription regulation.