Virtual filtering and screening of combinatorial libraries have recently gained attention as methods complementing the high-throughput screening and combinatorial chemistry. These chemoinformatic techniques rely heavily on quantitative structure-activity relationship (QSAR) analysis, a field with established methodology and successful history. In this review, we discuss the computational methods for building QSAR models. We start with outlining their usefulness in high-throughput screening and identifying the general scheme of a QSAR model. Following, we focus on the methodologies in constructing three main components of QSAR model, namely the methods for describing the molecular structure of compounds, for selection of informative descriptors and for activity prediction. We present both the well-established methods as well as techniques recently introduced into the QSAR domain.

BackgroundThe regulation of gene expression by transcription factors is a key determinant of cellular phenotypes. Deciphering genome-wide networks that capture which transcription factors regulate which genes is one of the major efforts towards understanding and accurate modeling of living systems. However, reverse-engineering the network from gene expression profiles remains a challenge, because the data are noisy, high dimensional and sparse, and the regulation is often obscured by indirect connections.ResultsWe introduce a gene regulatory network inference algorithm ENNET, which reverse-engineers networks of transcriptional regulation from a variety of expression profiles with a superior accuracy compared to the state-of-the-art methods. The proposed method relies on the boosting of regression stumps combined with a relative variable importance measure for the initial scoring of transcription factors with respect to each gene. Then, we propose a technique for using a distribution of the initial scores and information about knockouts to refine the predictions. We evaluated the proposed method on the DREAM3, DREAM4 and DREAM5 data sets and achieved higher accuracy than the winners of those competitions and other established methods.ConclusionsSuperior accuracy achieved on the three different benchmark data sets shows that ENNET is a top contender in the task of network inference. It is a versatile method that uses information about which gene was knocked-out in which experiment if it is available, but remains the top performer even without such information. ENNET is available for download from https://github.com/slawekj/ennet under the GNU GPLv3 license.

The NCI-60 cell line set is likely the most molecularly profiled set of human tumor cell lines in the world. However, a critical missing component of previous analyses has been the inability to place the massive amounts of "-omic" data in the context of functional protein signaling networks, which often contain many of the drug targets for new targeted therapeutics. We used reverse-phase protein array (RPPA) analysis to measure the activation/ phosphorylation state of 135 proteins, with a total analysis of nearly 200 key protein isoforms involved in cell proliferation, survival, migration, adhesion, etc., in all 60 cell lines. We aggregated the signaling data into biochemical modules of interconnected kinase substrates for 6 key cancer signaling pathways: AKT, mTOR, EGF receptor (EGFR), insulin-like growth factor-1 receptor (IGF-1R), integrin, and apoptosis signaling. The net activation state of these protein network modules was correlated to available individual protein, phosphoprotein, mutational, metabolomic, miRNA, transcriptional, and drug sensitivity data. Pathway activation mapping identified reproducible and distinct signaling cohorts that transcended organ-type distinctions. Direct correlations with the protein network modules involved largely protein phosphorylation data but we also identified direct correlations of signaling networks with metabolites, miRNA, and DNA data.

We propose a new classification method for the prediction of drug properties, called random feature subset boosting for linear discriminant analysis (LDA). The main novelty of this method is the ability to overcome the problems with constructing ensembles of linear discriminant models based on generalized eigenvectors of covariance matrices. Such linear models are popular in building classification-based structure-activity relationships. The introduction of ensembles of LDA models allows for an analysis of more complex problems than by using single LDA, for example, those involving multiple mechanisms of action. Using four data sets, we show experimentally that the method is competitive with other recently studied chemoinformatic methods, including support vector machines and models based on decision trees. We present an easy scheme for interpreting the model despite its apparent sophistication. We also outline theoretical evidence as to why, contrary to the conventional AdaBoost ensemble algorithm, this method is able to increase the accuracy of LDA models.

Inspection of structure changes in proteins borne by altering their sequences brings understanding of physics, functioning and evolution of existing proteins, and helps engineer modified ones. On single amino acid substitutions, the most frequent mutation type, shifts in backbone conformation are typically small, raising doubts if and how such minor modifications could drive evolutionary divergence. Here, we report that the distribution of magnitudes of structure change on such substitutions is heavy-tailed--whereas protein structures are robust to most substitutions, changes much larger than average occur with raised odds compared to what would be expected for exponential distribution with the same mean. This nonexponential behavior allows for reconciling the apparent contradiction between the observed conservation of protein structures and the substantial evolutionary plasticity implied in their diversity. The presence of the heavy tail in the distribution promotes structure divergence, facilitating exploration of new functionality, and conformations within folds, as well as exploration of structure space for new folds.

Network analysis has the potential to considerably contribute to the better understanding of the molecular mechanisms of wound healing and to the discovery of means to control and optimize that process.

BackgroundSomatic mutations accumulate in human cells throughout life. Some may have no adverse consequences, but some of them may lead to cancer. A cancer genome is typically unstable, and thus more mutations can accumulate in the DNA of cancer cells. An ongoing problem is to figure out which mutations are drivers - play a role in oncogenesis, and which are passengers - do not play a role. One way of addressing this question is through inspection of somatic mutations in DNA of cancer samples from a cohort of patients and detection of patterns that differentiate driver from passenger mutations.ResultsWe propose QuaDMutEx, a method that incorporates three novel elements: a new gene set penalty that includes non-linear penalization of multiple mutations in putative sets of driver genes, an ability to adjust the method to handle slow- and fast-evolving tumors, and a computationally efficient method for finding gene sets that minimize the penalty, through a combination of heuristic Monte Carlo optimization and exact binary quadratic programming. Compared to existing methods, the proposed algorithm finds sets of putative driver genes that show higher coverage and lower excess coverage in eight sets of cancer samples coming from brain, ovarian, lung, and breast tumors.ConclusionsSuperior ability to improve on both coverage and excess coverage on different types of cancer shows that QuaDMutEx is a tool that should be part of a state-of-the-art toolbox in the driver gene discovery pipeline. It can detect genes harboring rare driver mutations that may be missed by existing methods. QuaDMutEx is available for download from https://github.com/bokhariy/QuaDMutEx under the GNU GPLv3 license.

Background: Cancer is caused by genetic mutations, but not all somatic mutations in human DNA drive the emergence or growth of cancers. While many frequently-mutated cancer driver genes have already been identified and are being utilized for diagnostic, prognostic, or therapeutic purposes, identifying driver genes that harbor mutations occurring with low frequency in human cancers is an ongoing endeavor. Typically, mutations that do not confer growth advantage to tumors-passenger mutations-dominate the mutation landscape of tumor cell genome, making identification of low-frequency driver mutations a challenge. The leading approach for discovering new putative driver genes involves analyzing patterns of mutations in large cohorts of patients and using statistical methods to discriminate driver from passenger mutations. Results: We propose a novel cancer driver gene detection method, QuaDMutNetEx. QuaDMutNetEx discovers cancer drivers with low mutation frequency by giving preference to genes encoding proteins that are connected in human protein-protein interaction networks, and that at the same time show low deviation from the mutual exclusivity pattern that characterizes driver mutations occurring in the same pathway or functional gene group across a cohort of cancer samples. Conclusions: Evaluation of QuaDMutNetEx on four different tumor sample datasets show that the proposed method finds biologically-connected sets of low-frequency driver genes, including many genes that are not found if the network connectivity information is not considered. Improved quality and interpretability of the discovered putative driver gene sets compared to existing methods shows that QuaDMutNetEx is a valuable new tool for detecting driver genes. QuaDMutNetEx is available for download from https://github.com/bokhariy/QuaDMutNetEx under the GNU GPLv3 license.