Computational chemoproteomics to understand the role of selected psychoactives in treating mental health indications Jonathan Fine∗ Rachel Lackner† Ram Samudrala§¶ Gaurav Chopra∗‡¶ ∗ Department of Chemistry, Purdue University, West Lafayette, IN, USA † Yale University, New Haven, CT, USA ‡ Purdue Institute for Drug Discovery, Purdue Institute for Integrative Neuroscience, Purdue Institute for Integrative Neuroscience, Purdue Institute for Immunology, Inflammation and Infectious Disease, Purdue Center for Cancer Research, West Lafayette, IN, USA § Department of Biomedical Informatics, SUNY, Buffalo, NY, USA ¶ Corresponding authors: [email protected], [email protected] Abstract We have developed the Computational Analysis of Novel Drug Opportunities (CANDO) platform to infer homology of drug behavior at a proteomic level by constructing and analyzing structural compound-proteome interaction signatures of 3,733 compounds with 48,278 proteins in a shotgun manner. We applied the CANDO platform to predict putative therapeutic properties of 428 psychoactive compounds that belong to phenylethylamine, tryptamine, and cannabinoid chemical classes for treating mental health indications. Our findings indicate that these 428 psychoactives are among the top-ranked predictions for a significant fraction of mental health indications, demonstrating a significant preference for treating such indications over non-mental health indications, relative to randomized controls (p-value < 10-12). Also, we analyzed the use of specific tryptamines for the treatment of sleeping disorders, bupropion for substance abuse disorders, and cannabinoids for epilepsy. Our innovative use of the CANDO platform may guide the identification and development of novel therapies for mental health indications and provide an understanding of their causal basis on a detailed mechanistic level. These predictions can be used to provide new leads for preclinical drug development for mental health and other neurological disorders. Keywords: shotgun drug discovery, CANDO, mental health, psychoactive, multiscale modelling, computational docking Introduction Traditional approaches to drug discovery are highly specific to single targets and focus on limited sets of interactions between individual protein targets and small molecule compounds. Yet, the resulting treatments are applied universally to all patients. The goal of this approach is to completely inhibit the function of a protein responsible for pathogenesis and subsequently determine the compound’s toxicity or side effect profile for human use. Almost all current drugs have been developed by this approach; however, the number of novel drugs being discovered every year has been reduced to a handful; currently, there are less than 30 new drugs approved each year and most of them are analogs to existing drugs or are patent workarounds (data obtained from fda.gov). The estimated average costs for developing a novel drug and bringing it to market can be up to $2.6 billion, according to the Tufts Center for the Study of Drug Development <csdd.tufts.edu>. Thus, there exists a dearth of novel drug development, as the current process is both time and cost prohibitive 1–4. One solution is to repurpose/reposition existing drugs with relatively benign side effects for new indications 3,5–11. Systematic exploration of drug repurposing opportunities is hindered by extensive competition in the pharmaceutical industry, but computational approaches enable a workaround. We initially developed a drug repurposing approach for malaria based on computational multi-target docking with dynamics7, and since then we have validated our predictive models numerous times experimentally 3,8,9,12–17. We utilized this repurposing paradigm along with a computational platform that evaluates relationships between compound-proteome (chemoproteome) interaction signatures to predict proteome- and indication-specific drug regimens in a shotgun and holistic manner (i.e., against all indications simultaneously) 9,15. Here, we describe the application of our platform to identify possible therapeutic uses of 3 phenethylamines, tryptamines, and cannabinoids in treating/curing mental health indications. Leveraging computational chemoproteomics for drug discovery A majority of small molecule drugs are derived from plant sources 18–20, and since these molecules have risen as a result of a dynamic interplay of evolution between plants and other organisms sharing their environment, we hypothesize that functional small molecules that become drugs have multiple modes of action. We have thus developed a platform that is agnostic to the protein-compound interaction prediction (whether predicted or observed) and instead relies on whole "signatures of interactions”, which is either a binary or real value row of numbers representing how well a compound binds to a library of protein structures considered representative of the (current) structural universe. The platform assumes the similarity of compound-proteome interaction signatures as indicative of similar functional behavior, and non- similar signatures (or regions of signatures) are indicative of off- and anti-target (side) effects, in effect inferring homology of compound/drug behavior at a proteomic level. The signatures are used to rank compounds for all indications to provide an optimized and enriched set of verified protein-compound interactions, a comprehensive list of indications and compounds that may be readily repurposed, as well as mechanistic understanding of drug behavior at an atomic level. CANDO: A shotgun computational chemoproteomics platform for drug repurposing and discovery Biologically active molecules such as proteins and drugs do not function in isolation. The absorption, dispersion, metabolism, and excretion (ADME) and effectiveness of a drug are dependent on the interaction of the drugs with a system of proteins expressed at different sites in an organism. The Computational Analysis of Novel Drug Opportunities (CANDO) platform works at the proteomic level by leveraging the interaction signature of a compound to all proteins in a generic structural library, and compares the signatures of candidate compounds/drugs to those 4 approved for particular indications to make drug repurposing predictions in a shotgun manner (here meaning an all vs. all compound-proteome signature comparison). The first version of the CANDO platform shown in Figure 1 predicts interactions between 3,733 FDA approved drugs and a variety of other human ingestible compounds (including supplements and illegal drugs) and 48,278 protein structures (46,784 of which are used in this study) either taken from the Protein Data Bank 21 or representing a high confidence homology model22. We employ a hierarchical fragment-based docking with dynamics algorithm using knowledge-based potentials23 to construct a 3,733 x 48,278 compound-protein interaction matrix. We have previously shown that all-atom dynamics is necessary for accurate prediction of binding energies24 and demonstrated all-atom knowledge-based force fields are more accurate than physics-based approaches for both protein structure prediction and docking 16,17,25–28, and have shown that multi-targeted docking with dynamics leads to improved hit rates for finding inhibitors of pathogens relative to conventional approaches 7,8. Once the interaction matrix is constructed, our methods compares the all compound- proteome interaction signatures where the similarity of two signatures can be calculated using various metrics as simple as root mean squared deviations (RMSD) to sophisticated graph theory based comparisons that take underlying protein-protein interactions (compiled from public sources 22,29–31) into account. As mentioned above, similarities between (regions of) interaction signatures indicate a relationship in functional behavior. However, the differences between two signatures are difficult to understand without further knowledge as it may indicate a more potent drug, a possible side effect, or no effect whatsoever. In addition to producing a ranked list of putative drugs that are most likely to function similar to other drugs approved for a particular indication, the signature comparison and ranking helps to analyze compound behavior in biologically relevant 5 pathways29,30,32. Our CANDO platform is successful for prospectively validating putative leads for several indications 15,33,34. Mental health indications and interventions A large number of diseases and disorders have mental health implications as cataloged by the American Psychiatric Association (APA)35. These indications affect people in all age groups, social classes, and races 36–40. The treatments for these indications mostly consist of small molecule therapeutics, varying individually for specific diseases, disorders, or conditions. According to a report published by the World Health Organization in 2011 41, the number of United States (US) citizens taking medication to treat mental health has increased to over two million US citizens since 2001. Anxiety disorders make up the largest category of mental illness in the US affecting a total of 42 million people. The second largest category is major depression disorder affecting 14.8 million US citizens on any given day. Approximately 2.4 million US citizens have schizophrenia where no effective treatment or cure is currently available as schizophrenia medication typically results in metabolic issues leading to weight gain and type 2 diabetes42. Collectively mental