Pharmacology over the past 100 years has had a rich tradition of scientists with the ability to form qualitative or semiquantitative relations between molecular structure and activity in cerebro. To test these hypotheses they have consistently used traditional pharmacology tools such as in vivo and in vitro models. Increasingly over the last decade however we have seen that computational (in silico) methods have been developed and applied to pharmacology hypothesis development and testing. These in silico methods include databases, quantitative structure-activity relationships, pharmacophores, homology models and other molecular modeling approaches, machine learning, data mining, network analysis tools and data analysis tools that use a computer. In silico methods are primarily used alongside the generation of in vitro data both to create the model and to test it. Such models have seen frequent use in the discovery and optimization of novel molecules with affinity to a target, the clarification of absorption, distribution, metabolism, excretion and toxicity properties as well as physicochemical characterization. The aim of this review is to illustrate some of the in silico methods for pharmacology that are used in drug discovery. Further applications of these methods to specific targets and their limitations will be discussed in the second accompanying part of this review.

A variety of machine learning methods such as Naïve Bayesian, support vector machines and more recently deep neural networks are demonstrating their utility for drug discovery and development. These leverage the generally bigger data sets created from high throughput screening data and allow prediction of bioactivities for targets and molecular properties with increased levels of accuracy. We have only just begun to exploit the potential of these techniques but they may already be fundamentally changing the research process for identifying new molecules and/or repurposing old drugs. The integrated application of such machine learning models for end-to-end (E2E) application is broadly relevant and has considerable implications for developing future therapies and their targeting. Learning from history 'Those who do not remember the past are condemned to repeat it' (Santayana). This observation applies as much to drug discovery as it does to other aspects of human endeavor 1. The history of drug discovery is a prelude to the emerging potential of computerassisted data exploration. One constant in drug discovery is that every few years the estimated cost to develop drugs rises further. Less than 20 years ago, developing a drug took ~12 years, cost under a billion dollars, and the biggest challenges were failures due to efficacy or toxicity-induced attrition 2. in vitro pharmacological profiling implemented earlier in the drug discovery process helped to identify some predictable undesirable off-*

Machine learning methods have been applied to many datasets in pharmaceutical research for several decades. The relative ease and availability of fingerprint type molecular descriptors paired with Bayesian methods resulted in the widespread use of this approach for a diverse array of endpoints relevant to drug discovery. Deep learning is the latest machine learning algorithm attracting attention for many of pharmaceutical applications from docking to virtual screening. Deep learning is based on an artificial neural network with multiple hidden layers and has found considerable traction for many artificial intelligence applications. We have previously suggested the need for a comparison of different machine learning methods with deep learning across an array of varying datasets that is applicable to pharmaceutical research. Endpoints relevant to pharmaceutical research include absorption, distribution, metabolism, excretion and toxicity (ADME/Tox) properties, as well as activity against pathogens and drug discovery datasets. In this study, we have used datasets for solubility, probe-likeness, hERG, KCNQ1, bubonic plague, Chagas, tuberculosis and malaria to compare different machine learning methods using FCFP6 fingerprints. These datasets represent whole cell screens, individual proteins, physicochemical properties as well as a dataset with a complex endpoint. Our aim was to assess whether deep learning offered any improvement in testing when assessed using an array of metrics including AUC, F1 score, Cohen’s kappa, Matthews correlation coefficient and others. Based on ranked normalized scores for the metrics or datasets Deep Neural Networks (DNN) ranked higher than SVM, which in turn was ranked higher than all the other machine learning methods. Visualizing these properties for training and test sets using radar type plots indicates when models are inferior or perhaps over trained. These results also suggest the need for assessing deep learning further using multiple metrics with much larger scale comparisons, prospective testing as well as assessment of different fingerprints and DNN architectures beyond those used.

The protein product of the human ether-a-go-go gene (hERG) is a potassium channel that when inhibited by some drugs may lead to cardiac arrhythmia. Previously, a three-dimensional quantitative structure-activity relationship (3D-QSAR) pharmacophore model was constructed using Catalyst with in vitro inhibition data for antipsychotic agents. The rationale of the current study was to use a combination of in vitro and in silico technologies to further test the pharmacophore model and qualitatively predict whether molecules are likely to inhibit this potassium channel. These predictions were assessed with the experimental data using the Spearman's rho rank correlation. The antipsychotic-based hERG inhibitor model produced a statistically significant Spearman's rho of 0.71 for 11 molecules. In addition, 15 molecules from the literature were used as a further test set and were also well ranked by the same model with a statistically significant Spearman's rho value of 0.76. A Catalyst General hERG pharmacophore model was generated with these literature molecules, which contained four hydrophobic features and one positive ionizable feature. Linear regression of log-transformed observed versus predicted IC 50 values for this training set resulted in an r 2 value of 0.90. The model based on literature data was evaluated with the in vitro data generated for the original 22 molecules (including the antipsychotics) and illustrated a significant Spearman's rho of 0.77. Thus, the Catalyst 3D-QSAR approach provides useful qualitative predictions for test set molecules. The model based on literature data therefore provides a potentially valuable tool for discovery chemistry as future molecules may be synthesized that are less likely to inhibit hERG based on information provided by a pharmacophore for the inhibition of this potassium channel.

Bacterial survival requires an intact peptidoglycan layer, a 3-dimensional exoskeleton that encapsulates the cytoplasmic membrane. Historically, the final steps of peptidoglycan synthesis are known to be carried out by d,d-transpeptidases, enzymes that are inhibited by the β-lactams which constitute >50% of all antibacterials in clinical use. Here, we show that the carbapenem subclass of β-lactams is distinctly effective not only because they inhibit d,d-transpeptidases and are poor substrates for β-lactamases, but primarily because they also inhibit non-classical transpeptidases, namely the l,d-transpeptidases, that generate the majority of linkages in the peptidoglycan of mycobacteria. We have characterized the molecular mechanisms responsible for inhibition of l,d-transpeptidases of M. tuberculosis and a range of bacteria, including ESKAPE pathogens, and utilized this information to design, synthesize and test simplified carbapenems with potent antibacterial activity.