Antifreeze protein (AFP) is an ice-binding protein that protects organisms from freezing in extremely cold environments. AFPs are found across a diverse range of species and, therefore, significantly differ in their structures. As there are no consensus sequences available for determining the ice-binding domain of AFPs, thus the prediction and characterization of AFPs from their sequence is a challenging task. This study addresses this issue by predicting AFPs directly from sequence on a large set of 478 AFPs and 9,139 non-AFPs using machine learning (e.g., random forest) as a function of interpretable features (e.g., amino acid composition, dipeptide composition, and physicochemical properties). Furthermore, AFPs were characterized using propensity scores and important physicochemical properties via statistical and principal component analysis. The predictive model afforded high performance with an accuracy of 88.28% and results revealed that AFPs are likely to be composed of hydrophobic amino acids as well as amino acids with hydroxyl and sulfhydryl side chains. The predictive model is provided as a free publicly available web server called CryoProtect for classifying query protein sequence as being either AFP or non-AFP. The data set and source code are for reproducing the results which are provided on GitHub.

Introduction: Epigenetic modification has been implicated in a wide range of diseases and the ability to modulate such systems is a lucrative therapeutic strategy in drug discovery. Areas covered: This article focuses on the concepts and drug discovery aspects of epigenomics. This is achieved by providing a survey of the following concepts: (i) factors influencing epigenetics, (ii) diseases arising from epigenetics, (iii) epigenetic enzymes as druggable targets along with coverage of existing FDA-approved drugs and pharmacological agents, and (iv) drug repurposing/repositioning as a means for rapid discovery of pharmacological agents targeting epigenetics. Expert opinion: Despite significant interests in targeting epigenetic modifiers as a therapeutic route, certain classes of target proteins are heavily studied while some are less characterized. Thus, such orphan target proteins are not yet druggable with limited report of active modulators. Current research points towards a great future with novel drugs directed to the many complex multifactorial diseases of humans, which are still often poorly understood and difficult to treat.

Aim: Toxicity arising from hemolytic activity of peptides hinders its further progress as drug candidates. Materials & methods: This study describes a sequence-based predictor based on a random forest classifier using amino acid composition, dipeptide composition and physicochemical descriptors (named HemoPred). Results: This approach could outperform previously reported method and typical classification methods (e.g., support vector machine and decision tree) verified by fivefold cross-validation and external validation with accuracy and Matthews correlation coefficient in excess of 95% and 0.91, respectively. Results revealed the importance of hydrophobic and Cys residues on alpha-helix and beta-sheet, respectively, on the hemolytic activity. Conclusion: A sequence-based predictor which is publicly available as the web service of HemoPred, is proposed to predict and analyze the hemolytic activity of peptides.

Melanocortin signalling in leucocyte subsets elicits anti-inflammatory and immune tolerance inducing effects in animal experimental inflammation. In man, however, the effects of melanocortin signalling in inflammatory conditions have scarcely been examined. We explored the differential reactions of melanocortin 1-5 receptors (MC1-5R) gene expressions in pathogenetic leucocyte subsets in rheumatoid arthritis (RA) to treatment with TNF- inhibitor adalimumab. Seven patients with active RA donated blood at start and at 3-month treatment. CD4(+) T helper (h) lymphocytes (ly), CD8(+) T cytotoxic (c) ly, CD19(+) B ly and CD14(+) monocytes were isolated, using immunomagnetic beads, total RNA extracted and reverse transcription quantitative polymerase chain reaction (RT-qPCR) performed. Fold changes in MC1-5R, Th1-, inflammatory- and regulatory cytokine gene expressions were assessed for correlation. Six patients responded to adalimumab treatment, while one patient was non-responder. In all lymphocyte subtypes, MC1-5R gene expressions decreased in responders and increased in the non-responder. In responders, decrease in MC2R, MC3R and MC4R gene expressions in CD8(+) Tc and CD19(+) B ly was significant. Fold change in MC1-5R and IFN gene expressions correlated significantly in CD8(+) Tc ly, while fold change in MC1R, MC3R and MC5R and IL-1 gene expressions correlated significantly in CD4(+) Th ly. Our results show regulation of MC2R, MC3R and MC4R gene expressions in CD8(+) Tc ly and CD19(+) B ly. The correlations between fold change in different MCRs and disease driving cytokine gene expressions in CD8(+) Tc ly and CD4(+) Th ly point at a central immune modulating function of the melanocortin system in RA.

Host defense peptides (HDPs) represents a class of ubiquitous and rapid responding immune molecules capable of direct inactivation of a wide range of pathogens. Recent research has shown HDPs to be promising candidates for development as a novel class of broad-spectrum chemotherapeutic agent that is effective against both pathogenic microbes and malignant neoplasm. This study aims to quantitatively explore the relationship between easy-to-interpret amino acid composition descriptors of HDPs with their respective bioactivities. Classification models were constructed using the C4.5 decision tree and random forest classifiers. Good predictive performance was achieved as deduced from the accuracy, sensitivity and specificity in excess of 90% and Matthews correlation coefficient in excess of 0.5 for all three evaluated data subsets (e.g. training, 10-fold cross-validation and external validation sets). The source code and data set used for the construction of classification models are available on GitHub at https://github.com/chaninn/pepbio/.

Drug metabolism determines the fate of a drug when it enters the human body and is a critical factor in defining their absorption, distribution, metabolism, excretion and toxicity (ADMET) characteristics. Among the various drug metabolizing enzymes, cytochrome P450s (CYP450) constitute an important protein family that aside from functioning in xenobiotic metabolism is also responsible for a diverse array of other roles encompassing steroid and cholesterol biosynthesis, fatty acid metabolism, calcium homeostasis, neuroendocrine functions and growth regulation. Although CYP450 typically convert xenobiotics into safe metabolites, there are some situations whereby the metabolite is more toxic than its parent molecule. Computational modeling has been instrumental in CYP450 research by rationalizing the nature of the binding event (i.e. inhibit or induce CYP450s) or metabolic stability of query compounds of interest. A plethora of computational approaches encompassing ligand, structure and systems based approaches have been utilized to model CYP450-ligand interactions. This review provides a brief background on the CYP450 family (i.e. its roles, advantages and disadvantages as well as its modulators) and then discusses the various computational approaches that have been used to model CYP450-ligand interaction. Particular focus is given to the use of quantitative structure-activity relationship (QSAR) and more recent proteochemometric modeling studies. Finally, a perspective on the current state of the art and future trends of the field is provided.

Contamination of ground water by industrial chemicals presents a major environmental and health problem. Soil sorption plays an important role in the transport and movement of such pollutant chemicals. In this study, proteochemometric (PCM) modeling was used to unravel the origins of interactions of 17 phthalic acid esters (PAEs) against 3 soil types by predicting the organic carbon content normalized sorption coefficient (logK(oc)) values as a function of fingerprint descriptors of 17 PAEs and physical and textural properties of 3 soils. The results showed that PCM models provided excellent predictivity (R-2 = 0.94, Q(2) = 0.89,Q(Ext)(2) = 0.85). In further validation of the model, our proposed PCM model was assessed by leave-one-compound-out (Q(LOCO)(2) = 0.86) and leave-one-soil-out (Q(LOCO)(2) = 0.86) cross-validations. The transparency of the PCM model allowed interpretation of the underlying importance of descriptors, which potentially contributes to a better understanding on the outcome of PAEs in the environment. A thorough analysis of descriptor importance revealed the contribution of secondary carbon atoms on the hydrophobicity and flexibility of PAEs as significant properties in influencing the soil sorption capacity.

The increasing size of datasets in drug discovery makes it challenging to build robust and accurate predictive models within a reasonable amount of time. In order to investigate the effect of dataset sizes on predictive performance and modelling time, ligand-based regression models were trained on open datasets of varying sizes of up to 1.2 million chemical structures. For modelling, two implementations of support vector machines (SVM) were used. Chemical structures were described by the signatures molecular descriptor. Results showed that for the larger datasets, the LIBLINEAR SVM implementation performed on par with the well-established libsvm with a radial basis function kernel, but with dramatically less time for model building even on modest computer resources. Using a non-linear kernel proved to be infeasible for large data sizes, even with substantial computational resources on a computer cluster. To deploy the resulting models, we extended the Bioclipse decision support framework to support models from LIBLINEAR and made our models of logD and solubility available from within Bioclipse.

Aromatase, the rate-limiting enzyme that catalyzes the conversion of androgen to estrogen, plays an essential role in the development of estrogen-dependent breast cancer. Side effects due to aromatase inhibitors (AIs) necessitate the pursuit of novel inhibitor candidates with high selectivity, lower toxicity and increased potency. Designing a novel therapeutic agent against aromatase could be achieved computationally by means of ligand-based and structure-based methods. For over a decade, we have utilized both approaches to design potential AIs for which quantitative structure-activity relationships and molecular docking were used to explore inhibitory mechanisms of AIs towards aromatase. However, such approaches do not consider the effects that aromatase variants have on different AIs. In this study, proteochemometrics modeling was applied to analyze the interaction space between AIs and aromatase variants as a function of their substructural and amino acid features. Good predictive performance was achieved, as rigorously verified by 10-fold cross-validation, external validation, leave-one-compound-out cross-validation, leave-one-protein-out cross-validation and Y-scrambling tests. The investigations presented herein provide important insights into the mechanisms of aromatase inhibitory activity that could aid in the design of novel potent AIs as breast cancer therapeutic agents.

Background: Currently, monomeric fluorescent proteins (FP) are ideal markers for protein tagging. The prediction of oligomeric states is helpful for enhancing live biomedical imaging. Computational prediction of FP oligomeric states can accelerate the effort of protein engineering efforts of creating monomeric FPs. To the best of our knowledge, this study represents the first computational model for predicting and analyzing FP oligomerization directly from the amino acid sequence. Results: After data curation, an exhaustive data set consisting of 397 non-redundant FP oligomeric states was compiled from the literature. Results from benchmarking of the protein descriptors revealed that the model built with amino acid composition descriptors was the top performing model with accuracy, sensitivity and specificity in excess of 80% and MCC greater than 0.6 for all three data subsets (e.g. training, tenfold cross-validation and external sets). The model provided insights on the important residues governing the oligomerization of FP. To maximize the benefit of the generated predictive model, it was implemented as a web server under the R programming environment. Conclusion: osFP affords a user-friendly interface that can be used to predict the oligomeric state of FP using the protein sequence. The advantage of osFP is that it is platform-independent meaning that it can be accessed via a web browser on any operating system and device. osFP is freely accessible at http://codes.bio/osfp/ while the source code and data set is provided on GitHub at https://github.com/chaninn/osFP/.