News TN-Grid: A Computing System for Discovering Causal Relationships among Human Genes to Improve Drug Repositioning

[StefanR5R]

The scientists of the TN-Grid project are about to publish an article about their method of discovery of causal relationships in genes, and how it can be used for medical applications. A draft of the article is openly available.

In this paper we present the approach, methods and results of the gene@home project. Developed within a collaboration of Trento University with FEM and IMEM-CNR, the project aims to expand gene networks using transcriptomic datasets with the support of voluntary computation on the TN-GRID platform based on the BOINC system. The project involved so far two thousand volunteers and thousands of computers with a current estimated power of 14 teraFLOPS. In particular for human data we intend to provide a public resource to navigate and combine the
results by expanding each single human transcript. Such resource can have a substantial impact on biological and medical research, as we make evident in two case studies.

Starting from a local gene network (LGN), based or hypothesized on previous biological knowledge, its expansion consists in a set of genes and a list of interactions which describe putative causal relationships with the genes in the LGN. The expansion is calculated on observational gene expression data, organized in a coherent normalized data matrix. Each
expansion requires a few days to be carried out, even with the use of the BOINC distributed computing platform, thus presenting two main problems: the results cannot be provided to the user in real time and any expansion represents a unique elaboration, very unlikely to be submitted again given that the number of possible gene combinations is exponential in the size of a genome (5,000 - 30,000 genes).
To overcome these problems, we are now adopting an approach called OneGenE which aims to expand each single gene in an organism. [...] we will create a public database containing the expansions for each gene in an organism, thereby offering the possibility of building any
LGN expansion in a very short time by combining the already calculated single expansions.

The two case studies here presented are focused on drug repositioning for two quite common pathologies, i.e. prostate cancer and cardiovascular diseases. Drug repositioning is an alternative approach for the discovery of new therapeutic opportunities for already approved medicines. Compared to traditional de novo drug development strategies, which have become increasingly expensive and time-consuming, this method, which relies on previous knowledge and speed up the approval procedure of the drug regulators, can represent a valuable approach. The biological opportunity of drug repositioning relies on one side on the fact that many diseases share common dysregulated pathways and proteins, and on the other side that medicines actually
perturb multiple targets (off-targets interactions).

The article discusses not only lots of the theoretical background and methodology, but also how the BOINC platform works for them as a "socia-technical" infrastructure — e.g. how a BOINC volunteer helped optimize the application, and how TN-Grid's computational throughput peaked during the Formula BOINC sprint of September 2019.

 

[Orange Kid]

It is good to know that all the time, energy and money people but into these projects result in some good, real conclusions.