Periodic Reporting for period 2 - RiPCoN (Rapid interaction profiling of 2019-nCoV for network-based deep drug-repurpose learning (DDRL))
Okres sprawozdawczy: 2021-04-01 do 2022-03-31
The goal of this project is to understand which human proteins are targeted by the virus and which human molecular networks are changed and how. This important information will lead to a much improved understanding of the biology of coronaviruses, including the more deadly SARS and MERS. More importantly, with the help of this network information we can then use artificial intelligence and deep neural networks to identify drugs that are already on the market that revert some of the changes the virus aims to make. Thereby it may be possible to find treatment options that do not require long and costly clinical safety testing and may therefore become available much quicker. Moreover, we expect that the interactions we find will help understand the acute and long-term symptoms of patients suffereing from COVID-19. By better understanding the molecular causes, we expect to help alleviate symptoms and help patients.
Afterwards interactome mapping was started such that initial interactome data were available within few months after project start. these were subsequently verified and confirmed. The inital network from this project was available in summer. In parallel we initiated a collaboration with partners in Toronto, Canada, and Boston, US to increase the coverage of the network. These partners are contibuting additional interactions and the validation of the integrated interacome dataset is ongoing.
The network shows a strong enrichment of proteins related to inflammation and viral life cycle, especially vesicle trafficking. The integration with available genetic data demonstrates that interactions and their local network neighborhood are linked to metabolic features related to COVID-19, namely glycogen and fatty acid metabolism, and to diseases of the immune system, metabolism and metabolic syndrome and to neurological phenotypes.
In addition to the SARS-CoV-2 – Human contactome mentioned above we also generated two more critical resources available to the scientific community: i) the pan-CoV-ORFeome collection from 7 coronaviridae and ii) the high-quality comprehensive pan-CoV-human interactome network map (manuscript in prep).
Importantly, the protein coding ORFs enable the expression of all viral proteins for subsequent use in protein microarray platform or ELISA-based screens for patient screening, identification of immunogenic epitopes and thus contributing to vaccine development.
On the computational drug repurposing aspect we focused our efforts in the use of small molecule bioactivity signatures, together with natural-language processing techniques to rank the likeliness of activity against SARS-CoV-2 of all those approved and experimental drugs that experts around the world had suggested to be potentially active. Indeed, we developed a tool to help prioritize these potential treatments, stratifying them according to the level of clinical evidence and suggested mechanism of action for the intended drugs. The rapid deployment of this tool was aimed at helping clinical researchers in their (almost blind) initial choices (Duran-Frigola et al. 2020 J Chem Inf Model). Additionally, as reported, we also implemented a collection of deep-networks to infer bioactivity signatures to any molecule of interest (i.e. the Signaturizers; Bertoni et al. 2021 Nat Commun).