Setting Our Sights on Infectious Diseases.

In May 2019, the Wellcome Centre for Anti-Infectives Research (WCAIR) at the University of Dundee, UK, held an international conference with the aim of discussing some key questions around discovering new medicines for infectious diseases and a particular focus on diseases affecting Low and Middle Income Countries. There is an urgent need for new drugs to treat most infectious diseases. We were keen to see if there were lessons that we could learn across different disease areas and between the preclinical and clinical phases with the aim of exploring how we can improve and speed up the drug discovery, translational, and clinical development processes. We started with an introductory session on the current situation and then worked backward from clinical development to combination therapy, pharmacokinetic/pharmacodynamic (PK/PD) studies, drug discovery pathways, and new starting points and targets. This Viewpoint aims to capture some of the learnings.

Classification and analysis of a large collection of in vivo bioassay descriptions.

Testing potential drug treatments in animal disease models is a decisive step of all preclinical drug discovery programs. Yet, despite the importance of such experiments for translational medicine, there have been relatively few efforts to comprehensively and consistently analyze the data produced by in vivo bioassays. This is partly due to their complexity and lack of accepted reporting standards-publicly available animal screening data are only accessible in unstructured free-text format, which hinders computational analysis. In this study, we use text mining to extract information from the descriptions of over 100,000 drug screening-related assays in rats and mice. We retrieve our dataset from ChEMBL-an open-source literature-based database focused on preclinical drug discovery. We show that in vivo assay descriptions can be effectively mined for relevant information, including experimental factors that might influence the outcome and reproducibility of animal research: genetic strains, experimental treatments, and phenotypic readouts used in the experiments. We further systematize extracted information using unsupervised language model (Word2Vec), which learns semantic similarities between terms and phrases, allowing identification of related animal models and classification of entire assay descriptions. In addition, we show that random forest models trained on features generated by Word2Vec can predict the class of drugs tested in different in vivo assays with high accuracy. Finally, we combine information mined from text with curated annotations stored in ChEMBL to investigate the patterns of usage of different animal models across a range of experiments, drug classes, and disease areas.

Rational design of non-resistant targeted cancer therapies.

Drug resistance is one of the major problems in targeted cancer therapy. A major cause of resistance is changes in the amino acids that form the drug-target binding site. Despite of the numerous efforts made to individually understand and overcome these mutations, there is a lack of comprehensive analysis of the mutational landscape that can prospectively estimate drug-resistance mutations. Here we describe and computationally validate a framework that combines the cancer-specific likelihood with the resistance impact to enable the detection of single point mutations with the highest chance to be responsible of resistance to a particular targeted cancer therapy. Moreover, for these treatment-threatening mutations, the model proposes alternative therapies overcoming the resistance. We exemplified the applicability of the model using EGFR-gefitinib treatment for Lung Adenocarcinoma (LUAD) and Lung Squamous Cell Cancer (LSCC) and the ERK2-VTX11e treatment for melanoma and colorectal cancer. Our model correctly identified the phenotype known resistance mutations, including the classic EGFR-T790M and the ERK2-P58L/S/T mutations. Moreover, the model predicted new previously undescribed mutations as potentially responsible of drug resistance. Finally, we provided a map of the predicted sensitivity of alternative ERK2 and EGFR inhibitors, with a particular highlight of two molecules with a low predicted resistance impact.

Computational and Practical Aspects of Drug Repositioning.

The concept of the hypothesis-driven or observational-based expansion of the therapeutic application of drugs is very seductive. This is due to a number of factors, such as lower cost of development, higher probability of success, near-term clinical potential, patient and societal benefit, and also the ability to apply the approach to rare, orphan, and underresearched diseases. Another highly attractive aspect is that the "barrier to entry" is low, at least in comparison to a full drug discovery operation. The availability of high-performance computing, and databases of various forms have also enhanced the ability to pose reasonable and testable hypotheses for drug repurposing, rescue, and repositioning. In this article we discuss several factors that are currently underdeveloped, or could benefit from clearer definition in articles presenting such work. We propose a classification scheme-drug repositioning evidence level (DREL)-for all drug repositioning projects, according to the level of scientific evidence. DREL ranges from zero, which refers to predictions that lack any experimental support, to four, which refers to drugs approved for the new indication. We also present a set of simple concepts that can allow rapid and effective filtering of hypotheses, leading to a focus on those that are most likely to lead to practical safe applications of an existing drug. Some promising repurposing leads for malaria (DREL-1) and amoebic dysentery (DREL-2) are discussed.

Mycobacterial dihydrofolate reductase inhibitors identified using chemogenomic methods and in vitro validation.

The lack of success in target-based screening approaches to the discovery of antibacterial agents has led to reemergence of phenotypic screening as a successful approach of identifying bioactive, antibacterial compounds. A challenge though with this route is then to identify the molecular target(s) and mechanism of action of the hits. This target identification, or deorphanization step, is often essential in further optimization and validation studies. Direct experimental identification of the molecular target of a screening hit is often complex, precisely because the properties and specificity of the hit are not yet optimized against that target, and so many false positives are often obtained. An alternative is to use computational, predictive, approaches to hypothesize a mechanism of action, which can then be validated in a more directed and efficient manner. Specifically here we present experimental validation of an in silico prediction from a large-scale screen performed against Mycobacterium tuberculosis (Mtb), the causative agent of tuberculosis. The two potent anti-tubercular compounds studied in this case, belonging to the tetrahydro-1,3,5-triazin-2-amine (THT) family, were predicted and confirmed to be an inhibitor of dihydrofolate reductase (DHFR), a known essential Mtb gene, and already clinically validated as a drug target. Given the large number of similar screening data sets shared amongst the community, this in vitro validation of these target predictions gives weight to computational approaches to establish the mechanism of action (MoA) of novel screening hit.