A multilayer network approach for guiding drug repositioning in neglected diseases

BERENSTEIN AJ; MAGARIÑOS MP; CHERNOMORETZ A; AGÜERO F

Belo Horizonte

Conferencia; 3rd International ISCB Latin America Conference; 2014

International Society for Computational Biology

Background. Neglected tropical diseases are human infectious diseases that occur in tropical or subtropical regions and are often associated with poverty. Historically, lack of interest from the pharmaceutical industry, resulted in the lack of drugs to combat the majority of the pathogens that cause these diseases. With the advent of public domain chemical resources and the open release of data from high throughput screening assays the availability of chemical information has increased greatly. However most of the information on targets and the activity of drugs is for humans or model organisms. This creates a huge opportunity for computational exercises to identify candidates for drug discovery in neglected diseases by using comparative genomics, and chemogenomic approaches. In this work we use a multilayer network strategy to model a map of bioactive drugs, their targets, and other infromative relationships. Materials and Methods. Chemical data-sets, including bioactivity data against pathogen and non-pathogen targets were obtained from open databases and high throughput screenings. Using these data, we built a three-layer network containing three disjoint sets of vertexes with 1.48 106 drugs and 1.67 105 proteins across 221 species and three affiliation-type protein features (orthology, Pfam domains, participation in metabolic pathways). Only statistically significant features (in the context of drug-target predictions) were taken into account. A bipartite projection was done over the protein layer. In the resulting monopartite protein projected network, proteins are linked only if they share at least one relevant biological entity. Using this model, we tackled the problems of i) prioritizing targets for drug discovery for pathogens; and ii) suggest candidate targets for orphan compounds (those that are active in whole-cell screenings but whose target is currently unknown). Results. Our approach allowed us to get statistically significant prioritized lists in both pathogen and model organisms, as evaluated by a tenfold cross validation procedure. Moreover, we found that our method overcomes traditional similarity (equence-alignment) based searches against druggable targets. After validating the method we applied it to prioritize Trypanosoma cruzi proteins. In the presentation we will discuss some of the top ranked proteins. We also used obtained candidate target proteins for orphan molecules that were bioactive in whole-cell assays against P. falciparum. We will discuss the proposed targets that mediate the antiplasmodial activity for some of the orphan compounds. Conclusions. We identified candidate drug targets, either for complete query species or for orphan compounds. Some of them were already validated and other are potentially new. The method applied allowed us to take advantage of the great amount of bioactivity data existent for model organisms, in order to make inferences concerning less studied pathogen species. Supported by: CONICET (fellowships and salaries), ANPCyT (Agencia Nacional de Promoción Científica y Tecnológica (grants PICT-2010-1479, PICTO-Glaxo-2013-0067).