Integrative relational machine-learning for understanding drug side-effect profiles.

BACKGROUND: Drug side effects represent a common reason for stopping drug development during clinical trials. Improving our ability to understand drug side effects is necessary to reduce attrition rates during drug development as well as the risk of discovering novel side effects in available drugs. Today, most investigations deal with isolated side effects and overlook possible redundancy and their frequent co-occurrence. RESULTS: In this work, drug annotations are collected from SIDER and DrugBank databases. Terms describing individual side effects reported in SIDER are clustered with a semantic similarity measure into term clusters (TCs). Maximal frequent itemsets are extracted from the resulting drug x TC binary table, leading to the identification of what we call side-effect profiles (SEPs). A SEP is defined as the longest combination of TCs which are shared by a significant number of drugs. Frequent SEPs are explored on the basis of integrated drug and target descriptors using two machine learning methods: decision-trees and inductive-logic programming. Although both methods yield explicit models, inductive-logic programming method performs relational learning and is able to exploit not only drug properties but also background knowledge. Learning efficiency is evaluated by cross-validation and direct testing with new molecules. Comparison of the two machine-learning methods shows that the inductive-logic-programming method displays a greater sensitivity than decision trees and successfully exploit background knowledge such as functional annotations and pathways of drug targets, thereby producing rich and expressive rules. All models and theories are available on a dedicated web site. CONCLUSIONS: Side effect profiles covering significant number of drugs have been extracted from a drug ×side-effect association table. Integration of background knowledge concerning both chemical and biological spaces has been combined with a relational learning method for discovering rules which explicitly characterize drug-SEP associations. These rules are successfully used for predicting SEPs associated with new drugs.

eTRIKS platform: Conception and operation of a highly scalable cloud-based platform for translational research and applications development.

We describe the genesis, design and evolution of a computing platform designed and built to improve the success rate of biomedical translational research. The eTRIKS project platform was developed with the aim of building a platform that can securely host heterogeneous types of data and provide an optimal environment to run tranSMART analytical applications. Many types of data can now be hosted, including multi-OMICS data, preclinical laboratory data and clinical information, including longitudinal data sets. During the last two years, the platform has matured into a robust translational research knowledge management system that is able to host other data mining applications and support the development of new analytical tools.

Molecular modeling and simulation of water near model micelles: diffusion, rotational relaxation and structure at the hydration interface.

This paper reports on molecular dynamics simulations of two hydrated micelles composed of C12E6 and LDAO surfactants. The simulations results provide a quantitative picture of the dynamics of the hydration water at the water/micelle interface. Both the residence time of water near the micelle surface and its retardation with respect to the bulk have been estimated. It is found that the water dynamics is radically different for the two micellar systems and depends on the physical nature of the micelle surface in contact with water. For C12E6 this interface is thicker and presents a stronger hydrophilic character than that of LDAO. Thus, in C12E6, surface water dynamics is 1-2 orders of magnitude slower than that of bulk water, compared with only 18% for the LDAO system. The simulations have also revealed the nature of the rotational landscape experienced by water at the micellar surface: In the C12E6 micelle water rotation occurs in a highly anisotropic space due to confinement of waters at the interface; in LDAO the rotational landscape is instead isotropic. These findings clearly indicate that the slowdown of interfacial water relaxation near complex micelles depends, case by case, on the structural properties of the interface itself, such as the ratio between hydrophobic/hydrophilic exposed regions and on the interface thickness and topography.

Benchmarking of HPCC: A novel 3D molecular representation combining shape and pharmacophoric descriptors for efficient molecular similarity assessments.

Since 3D molecular shape is an important determinant of biological activity, designing accurate 3D molecular representations is still of high interest. Several chemoinformatic approaches have been developed to try to describe accurate molecular shapes. Here, we present a novel 3D molecular description, namely harmonic pharma chemistry coefficient (HPCC), combining a ligand-centric pharmacophoric description projected onto a spherical harmonic based shape of a ligand. The performance of HPCC was evaluated by comparison to the standard ROCS software in a ligand-based virtual screening (VS) approach using the publicly available directory of useful decoys (DUD) data set comprising over 100,000 compounds distributed across 40 protein targets. Our results were analyzed using commonly reported statistics such as the area under the curve (AUC) and normalized sum of logarithms of ranks (NSLR) metrics. Overall, our HPCC 3D method is globally as efficient as the state-of-the-art ROCS software in terms of enrichment and slightly better for more than half of the DUD targets. Since it is largely admitted that VS results depend strongly on the nature of the protein families, we believe that the present HPCC solution is of interest over the current ligand-based VS methods.