Effective holistic characterization of small molecule effects using heterogeneous biological networks

Abstract

Drug discovery is the practice of identifying chemical entities with activities useful for treating human diseases. Despite substantial advancements in biotechnologies such as assays for detecting promising hits and computational methods for identifying targets to pursue, the success rate of novel therapeutics in the clinic is rapidly declining. The two most common reasons for drug attrition in clinical trials are efficacy, in that the drug is not capable of treating the disease, and safety, in which the compound is too toxic to patients and does not outweigh any purported benefits. Here we describe the use of an integrated human interactome network to describe the complete picture of drug behavior in biological systems and lead to more promising therapeutic candidate generation. The Computational Analysis of Novel Drug Opportunities (CANDO) platform for shotgun multiscale drug discovery, repurposing, and design was enhanced to include drugs/compounds, proteins, diseases/indications, Gene Ontology entities, and drug side effects, accompanied with various known associations between them, and interactomic signatures were generated for all compounds using the node2vec algorithm for the purpose of relating them in the context of diseases for which they are approved. The results indicate there is significant biological information available in the side effect profile of a drug and they can be used to better relate drugs in terms of their therapeutic behavior. Side effects were also predicted for all compounds in the platform; known drug-side effect associations were recovered at rates ten times higher than what can be expected with random chance. The network architecture was also investigated in the context of protein pathways, where drug distance to various pathways served as the features for input to a random forest machine learning model in the context of indications they are associated to treat, with interesting results highlighted for mental disorders and cancer metastasis. This pipeline highlights the ability of CANDO to accurately relate drugs in a multitarget interactomic context, and paves the way for predicting novel putative drug candidates using the information gleaned from not only their side effect profiles, but their impacts on protein pathways as well.

Patient-derived organoids for precision oncology: a platform to facilitate clinical decision making

“…The incidence of cancer is increasing globally, posing a significant challenge in developing effective and economically viable anticancer drugs. Clinical approval for drugs entering phase I clinical trials after preclinical studies is crucial, yet the current rate of clinical approval for potential anticancer drugs is as low as 5%, much lower than drugs for other diseases [ 1 ]. Identifying a safe, potent and efficacious drug requires thorough preclinical testing, which evaluates aspects of pharmacodynamics, pharmacokinetics, and toxicology in in vitro and in vivo settings [ 2 ].…”

Optimal COVID-19 therapeutic candidate discovery using the CANDO platform

“…CANDO was originally designed as a shotgun repurposing platform for exactly this type of epidemic/pandemic scenario utilizing multiscale modeling techniques and adhering to multitarget drug theory, but has since been enhanced to carry out novel drug discovery against all indications ( Jenwitheesuk and Samudrala, 2003b , 2005 ; Jenwitheesuk et al, 2008 ; Horst et al, 2012 ; Minie et al, 2014 ; Sethi et al, 2015 ; Chopra et al, 2016 ; Chopra and Samudrala, 2016 ; Falls et al, 2019 ; Fine et al, 2019 ; Mangione and Samudrala, 2019 ; Schuler et al, 2019 ; Schuler and Samudrala, 2019 ; Mangione et al, 2020b ; Hudson and Samudrala, 2021 ; Schuler et al, 2021 ) as well as novel drug design ( Overhoff et al, 2021 ). The relatively recent introduction of higher order biological data such as protein pathways, protein-protein interactions, drug side effects, and protein-disease associations has further augmented our ability to describe compound behavior holistically, with subsequent improved performance ( Moukheiber et al, 2021 ; Schuler et al, 2021 ; Mangione, 2022 ; Mangione et al, 2022 ). Our platform is freely available to the scientific community and a detailed description of the software implementation has been published ( Mangione et al, 2020a ).…”