ALOHA: a novel probability fusion approach for scoring multi-parameter drug-likeness during the lead optimization stage of drug discovery.

Automated lead optimization helper application (ALOHA) is a novel fitness scoring approach for small molecule lead optimization. ALOHA employs a series of generalized Bayesian models trained from public and proprietary pharmacokinetic, absorption, distribution, metabolism, and excretion, and toxicology data to determine regions of chemical space that are likely to have excellent drug-like properties. The input to ALOHA is a list of molecules, and the output is a set of individual probabilities as well as an overall probability that each of the molecules will pass a panel of user selected assays. In addition to providing a summary of how and when to apply ALOHA, this paper will discuss the validation of ALOHA's Bayesian models and probability fusion approach. Most notably, ALOHA is demonstrated to discriminate between members of the same chemical series with strong statistical significance, suggesting that ALOHA can be used effectively to select compound candidates for synthesis and progression at the lead optimization stage of drug discovery.

Navigating the kinome.

Although it is increasingly being recognized that drug-target interaction networks can be powerful tools for the interrogation of systems biology and the rational design of multitargeted drugs, there is no generalized, statistically validated approach to harmonizing sequence-dependent and pharmacology-dependent networks. Here we demonstrate the creation of a comprehensive kinome interaction network based not only on sequence comparisons but also on multiple pharmacology parameters derived from activity profiling data. The framework described for statistical interpretation of these network connections also enables rigorous investigation of chemotype-specific interaction networks, which is critical for multitargeted drug design.

Rational approaches to targeted polypharmacology: creating and navigating protein-ligand interaction networks.

Many successful drugs bind to and modulate multiple targets in vivo. Successfully navigating protein-ligand polypharmacology will be a crucial and increasingly utilized component of pharmaceutical research. As publicly available databases of ligand activity values continue to grow in size and quality, infrastructure is needed to enable scientists to create and interact with these networks to fuel hypothesis-driven science. While most of the individual tools for creating this infrastructure exist, effectively connecting the data to the network to the scientist is very much a work in progress. Standards for publishing network data are also important to facilitate the analysis and comparison of networks from different research groups using different methods.

Evolving molecules using multi-objective optimization: applying to ADME/Tox.

Modern drug discovery involves the simultaneous optimization of many physicochemical and biological properties that transcends the historical focus on bioactivity alone. The process of resolving many requirements is termed 'multi-objective optimization', and here we discuss how this can be used for drug discovery, focusing on evolutionary molecule design to incorporate optimal predicted absorption, distribution, metabolism, excretion and toxicity properties. We provide several examples of how Pareto optimization implemented in Pareto Ligand Designer can be used to make trade-offs between these different predicted or real molecular properties to result in better predicted compounds.

Application of belief theory to similarity data fusion for use in analog searching and lead hopping.

A wide variety of computational algorithms have been developed that strive to capture the chemical similarity between two compounds for use in virtual screening and lead discovery. One limitation of such approaches is that, while a returned similarity value reflects the perceived degree of relatedness between any two compounds, there is no direct correlation between this value and the expectation or confidence that any two molecules will in fact be equally active. A lack of a common framework for interpretation of similarity measures also confounds the reliable fusion of information from different algorithms. Here, we present a probabilistic framework for interpreting similarity measures that directly correlates the similarity value to a quantitative expectation that two molecules will in fact be equipotent. The approach is based on extensive benchmarking of 10 different similarity methods (MACCS keys, Daylight fingerprints, maximum common subgraphs, rapid overlay of chemical structures (ROCS) shape similarity, and six connectivity-based fingerprints) against a database of more than 150,000 compounds with activity data against 23 protein targets. Given this unified and probabilistic framework for interpreting chemical similarity, principles derived from decision theory can then be applied to combine the evidence from different similarity measures in such a way that both capitalizes on the strengths of the individual approaches and maintains a quantitative estimate of the likelihood that any two molecules will exhibit similar biological activity.

Topography-biased compound library design: the shape of things to come?

The design and synthesis of quality compound libraries is of critical importance to any pharmaceutical company that relies on high throughput screening efforts for the identification of lead compounds. In this perspective, we use a moment of inertia derived shape analysis to interrogate potential libraries for chemical synthesis. An analysis of known 'Rule of Five' compliant drug shapes using this methodology clearly highlights compound libraries that may be reasonably expected, shape wise, to interact with biologically relevant protein active site topography and those that, although being structurally diverse in shape, have little chance of being pharmacologically productive. The use of multicomponent reactions as a means of producing structurally novel, bioactive compounds in a synthetically expeditious manner is also highlighted.

Enhancement of chemical rules for predicting compound reactivity towards protein thiol groups.

Non-specific chemical modification of protein thiol groups continues to be a significant source of false positive hits from high-throughput screening campaigns and can even plague certain protein targets and chemical series well into lead optimization. While experimental tools exist to assess the risk and promiscuity associated with the chemical reactivity of existing compounds, computational tools are desired that can reliably identify substructures that are associated with chemical reactivity to aid in triage of HTS hit lists, external compound purchases, and library design. Here we describe a Bayesian classification model derived from more than 8,800 compounds that have been experimentally assessed for their potential to covalently modify protein targets. The resulting model can be implemented in the large-scale assessment of compound libraries for purchase or design. In addition, the individual substructures identified as highly reactive in the model can be used as look-up tables to guide chemists during hit-to-lead and lead optimization campaigns.