DeepTransfo

Horizontal gene transfer through natural transformation (NT) is a major driver of bacterial evolution. In many pathogenic species, it is the main mechanism for the spreading of antibiotic resistances and virulence factors. Despite the importance of NT in bacterial pathogenesis, many players and aspects of the underlying mechanisms remain to be unveiled. Natural transformation is a multi-step process initiated by the capture of exogenous double-stranded DNA molecules at the surface of bacterial cells by dedicated macromolecular complexes that deliver the transforming DNA (tDNA) into the periplasm/cell wall space (uptake). The transport of tDNA into the cytoplasm is carried out by the polytopic membrane protein ComEC together with cytosolic proteins that provide the driving force for its internalisation and handle it to the recombination machinery for integration into the bacterial chromosome. While the proteins involved in the natural transformation steps are in some cases different between species, the basic processes seem to be highly conserved. A major limitation so far in the characterisation of the uptake and transport steps of NT has been the lack of biochemical and structural information on the proteins involved, in particular for the membrane embedded ComEC. To overcome this obstacle, we propose to develop deep learning-based models of ComEC and the proteins connecting it to the uptake and recombination machineries to guide biochemical, proteomic and in vivo experiments (genetics, fluorescence microscopy) using two distant human pathogens, Helicobacter pylori and Streptococcus pneumoniae, as models. Back and forth exchanges between the novel deep learning based predictions by Partner 3 and the experiments conducted by Partners 1 and 2, specialists of H. pylori and S. pneumoniae natural transformation systems, respectively, will uncover the molecular mechanisms underlying natural transformation.