Discovery of novel peptide-GPCR signalling systems across evolutionary domains
Navn på bevillingshaver
Alexander Hauser
Titel
Associate Professor
Institution
University of Copenhagen
Beløb
DKK 4,799,049
År
2022
Bevillingstype
Semper Ardens: Accelerate
Hvad?
The genomic revolution has allowed researchers to generate and analyse genome-scale datasets to improve our understanding of fundamental ecological and evolutionary processes. However, little focus has been given on inter-species molecular interactions mediated by peptide:receptor couplings. Here, I want to develop state-of-the-art comparative genomics and end-to-end deep learning pipelines to find inter-species and across domains of life peptide:receptor pairs that have co-evolved for their respective ecological niche. This includes, but is not limited to, microbiome-derived peptide-hormones, plant-derived sweet-tasting proteins, and venom-derived peptides from cone-snails, snakes, and spiders, elucidating novel molecular interconnections between species retelling evolutionary processes.
Hvorfor?
Organisms constantly interact with their environment. At the molecular level, some of these interactions are driven directly or indirectly through small molecules or peptides. Among naturally occurring compounds, peptides from plants, cone-snails, snakes, spiders, fungi, and bacteria are of particular interest as they cover a chemical space directly encoded in the organisms’ genomes. Some of these peptides are known to interact across species such as for fertilization, communication, and defense mechanisms. While the genomic revolution has allowed researchers to generate and analyse genome-scale datasets to identify putative peptides, interactions to their respective receptors have mostly been discovered by chance. Here, we want to employ novel computational methods and datasets to discover and predict inter-species peptide:receptor interactions to explain how biophysical and functional relationships remain coordinated over evolutionary time.
Hvordan?
Deep learning-based approaches, such as AlphaFold2 (AF2), achieved remarkable performances on the prediction of protein structures. Moreover, these methods also outperform classical protein-protein docking strategies. This project will significantly expand this analytical framework, incorporating ever-increasing publicly available data including thousands of metagenomes from microbial communities, millions of structural protein models, and a plethora of omics-dataset from all clades of life. We will build and utilize end-to-end deep-learning protocols based on AF2’s transformative methods and models underlying residue-residue co-evolution.