Despite being a recurring type of sequence variation, amino acid insertions and deletions (InDels) and their resulting functional significance remain a rather unexplored area of structural biology. InDels are quite often the driving force behind many diseases. Despite that, modeling InDels and exploring their functional implications remains lacking, mainly due to the dearth of experimental information and computational methodologies. In this work we introduce an algorithmic approach to model short InDels in silico and explore the structural and rigidity differences between the wildtype and the mutant protein structures. We assess rigidity to gather useful information about the protein’s general structure, the location of flexible and rigid clusters, and to permit a visual analysis of the effects of InDels local to the mutation site. Our results show that our method can efficiently create a computer-generated mutant that is functionally similar to the experimental mutant at both the local region of the InDel, as well as on the entire protein scale. The results show promise in our ability to accurately predict the effects of short insertions and deletions on the structural properties of proteins.

MOTIVATION: Over the past decade, there have been impressive advances in determining the 3D structures of protein complexes. However, there are still many complexes with unknown structures, even when the structures of the individual proteins are known. The advent of protein sequence information provides an opportunity to leverage evolutionary information to enhance the accuracy of protein-protein interface prediction. To this end, several statistical and machine learning methods have been proposed. In particular, direct coupling analysis has recently emerged as a promising approach for identification of protein contact maps from sequential information. However, the ability of these methods to detect protein-protein inter-residue contacts remains relatively limited. RESULTS: In this work, we propose a method to integrate sequential and co-evolution information with structural and functional information to increase the performance of protein-protein interface prediction. Further, we present a post-processing clustering method that improves the average relative F1 score by 70% and 24% and the average relative precision by 80% and 36% in comparison with two state-of-the-art methods, PSICOV and GREMLIN. AVAILABILITY AND IMPLEMENTATION: https://github.com/BioMLBoston/PatchDCA. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Understanding the nature of many diseases, including cancer, requires locating somatically acquired rearrangements corresponding to large-scale chromosomal aberrations. Computational methods to detect inter-chromosomal rearrangements based on next-generation sequencing platforms face the big challenge of accurately predicting the location of sites spanned by a typically small number of reads, while the entire sample contains hundreds of millions of reads. In this work, we propose a method called TDJD that identifies the location of interchromosomal breakpoints corresponding to a large scale structural variations, in particular translocations and insertions. To reduce the huge dimension of the search space, we split candidate reads that can be potential break points into windows, and represent the windows as a sequence of binary fingerprints. We then search for the location of the breakpoint in the reference genome using Jaccard distance. We use a combination of parallel computing, search using Jaccard distance to solve the exact nearest neighbor problem. The dimensionality reduction takes advantage of an SSE multi-thread architecture to achieve efficient search. We applied our algorithm to identify several reads with breakpoints, including those characterizing the PAX8-PPARγ rearrangement, a frequent modification occurring in follicular thyroid cancer. Our results show that we could identify the breakpoints much faster than the previous method. We also compared our results to several recently published methods, and found that our method is faster than all other compared methods with high accuracy.

BACKGROUND: Understanding protein structure and dynamics is essential for understanding their function. This is a challenging task due to the high complexity of the conformational landscapes of proteins and their rugged energy levels. In particular, it is important to detect highly populated regions which could correspond to intermediate structures or local minima. RESULTS: We present a hierarchical clustering and algebraic topology based method that detects regions of interest in protein conformational space. The method is based on several techniques. We use coarse grained protein conformational search, efficient robust dimensionality reduction and topological analysis via persistent homology as the main tools. We use two dimensionality reduction methods as well, robust Principal Component Analysis (PCA) and Isomap, to generate a reduced representation of the data while preserving most of the variance in the data. CONCLUSIONS: Our hierarchical clustering method was able to produce compact, well separated clusters for all the tested examples.

One of the major challenges for protein-protein docking is to accurately discriminate native-like structures from false-positives. While there is an agreement on the existence of a relationship between various favorable intermolecular interactions (e.g., Van der Waals, electrostatic, desolvation forces, etc.) and the similarity of a conformation to its native structure, the exact nature of this relationship is not clear. Different docking algorithms often formulate this relationship as a weighted sum of selected terms and calibrate their weights against a training set to evaluate and rank candidate complexes. Despite improvement in the predictive abilities of recent docking methods, even state-of-the-art methods often fail to predict the binding of many complexes and still output a large number of false positive complexes. We propose a novel machine learning approach that not only ranks candidate structures relative to each other, but also predicts how similar each candidate is to the native conformation. We trained a two-layer neural network, a deep neural network and a network of Restricted Boltzmann Machines against extensive datasets of unbound complexes. We tested these methods with a set of candidate structures. Our method is able to predict the RMSDs of unbound docked complexes with a very small, often <; 1.5Å error margin.

Microarray gene expression data comes as a time series, where the expression level of a gene is recorded at specific time points. Comparing the time series produced by two genes can give us information about the regulatory or inhibitory relationship between the genes. We present a Dynamic Programming (DP) method to compare gene expression data using geometric similarity. We aim to detect similarities and relationships between genes, based on their expression time series. By representing the time series as polygons and compare them, we can find relationships that are not available when the two time series are compared point-by-point. We applied our algorithm on a dataset of 343 regulatory pairs from the alpha dataset and compared them to randomly generated pairs. Using an SVM classifier, we find the optimal similarity score that separates the regulatory dataset from the random pairs. Our results show that we can detect similar pairs better than simple Pearson correlation and we outperform many of the existing methods. This method is an ongoing approach, that can be applied to finding the similarity of any data that can convert to 2D polygon. In the future, we plan to introduce this method as a new classifier.

Microarray gene expression data comes as a time series, where the expression level of a gene is recorded at specific time points. Comparing the time series produced by two genes can give us information about the regulatory or inhibitory relationship between the genes. We present a Dynamic Programming (DP) method to compare gene expression data using geometric similarity. We aim to detect similarities and relationships between genes, based on their expression time series. By representing the time series as polygons and compare them, we can find relationships that are not available when the two time series are compared point-by-point. We applied our algorithm on a dataset of 343 regulatory pairs from the alpha dataset and compared them to randomly generated pairs. Using an SVM classifier, we find the optimal similarity score that separates the regulatory dataset from the random pairs. Our results show that we can detect similar pairs better than simple Pearson correlation and we outperform many of the existing methods. This method is an ongoing approach, that can be applied to finding the similarity of any data that can convert to 2D polygon. In the future, we plan to introduce this method as a new classifier.

Structural variations (SVs) are deletions, duplications and rearrangements of medium to large segments (>100 base pairs (bp)) of the genome. Such genomic mutations are often described as being the primary cause of many diseases, including cancer. Breakpoint detection using next-generation sequencing (NGS) platforms still remains an open problem since computational methods to detect SVs face the big challenge of accurately predicting the precise location of breakpoints, which are typically spanned by a very small number of reads among millions that are generated during the sequencing process. In this work, we propose a method called TIDE to identify reads from paired-end sequencing data containing inter-chromosomal translocation or insertion breakpoints, which are specific types of SVs involving different chromosomes. To achieve this, we use discordant read pairs to narrow the search space and split prospective breakpoint-spanning reads into windows that are subsequently represented by a sequence of k-mers indexes, which we call fingerprints. We then apply a distance-preserving embedding algorithm to solve the approximate nearest neighbor problem of pairing the most similar fingerprints originated from the sample and reference genome. Experimental results show the efficacy of the method to find reads containing breakpoints characterizing the PAX8-PPARγ rearrangement found in thyroid cancer samples. We also compare our results with the ones provided by two recently published algorithms for detecting structural variation in clinical data.

MOTIVATION: Secondary-Structure Guided Superposition tool (SSGS) is a permissive secondary structure-based algorithm for matching of protein structures and in particular their fragments. The algorithm was developed towards protein structure prediction via fragment assembly. RESULTS: In a fragment-based structural prediction scheme, a protein sequence is cut into building blocks (BBs). The BBs are assembled to predict their relative 3D arrangement. Finally, the assemblies are refined. To implement this prediction scheme, a clustered structural library representing sequence patterns for protein fragments is essential. To create a library, BBs generated by cutting proteins from the PDB are compared and structurally similar BBs are clustered. To allow structural comparison and clustering of the BBs, which are often relatively short with flexible loops, we have devised SSGS. SSGS maintains high similarity between cluster members and is highly efficient. When it comes to comparing BBs for clustering purposes, the algorithm obtains better results than other, non-secondary structure guided protein superimposition algorithms.