Srinivas Somarowthu

Now on display: a gallery of group II intron structures at different stages of catalysis

Group II introns are mobile genetic elements that self-splice and retrotranspose into DNA and RNA. They are considered evolutionary ancestors of the spliceosome, the ribonucleoprotein complex essential for pre-mRNA processing in higher eukaryotes. Over a 20-year period, group II introns have been characterized first genetically, then biochemically, and finally by means of X-ray crystallography. To date, 17 crystal structures of a group II intron are available, representing five different stages of the splicing cycle. This review provides a framework for classifying and understanding these new structures in the context of the splicing cycle. Structural and functional implications for the spliceosome are also discussed.

POOL server

Summary: We present an automated web server for partial order optimum likelihood (POOL), a machine learning application that combines computed electrostatic and geometric information for high-performance prediction of catalytic residues from 3D structures. Input features consist of THEMATICS electrostatics data and pocket information from ConCavity. THEMATICS measures deviation from typical, sigmoidal titration behavior to identify functionally important residues and ConCavity identifies binding pockets by analyzing the surface geometry of protein structures. Both THEMATICS and ConCavity (structure only) do not require the query protein to have any sequence or structure similarity to other proteins. Hence, POOL is applicable to proteins with novel folds and engineered proteins. As an additional option for cases where sequence homologues are available, users can include evolutionary information from INTREPID for enhanced accuracy in site prediction. Availability: The web site is free and open to all users with no login requirements at http://www.pool.neu.edu. Contact: m.ondrechen@neu.edu Supplementary Information: Supplementary data are available at Bioinformatics online.

High-performance prediction of functional residues in proteins with machine learning and computed input features†

One of the major challenges in genomics is to understand the function of gene products from their 3D structures. Computational methods are needed for the high-throughput prediction of the function of proteins from their 3D structure. Methods that identify active sites are important for understanding and annotating the function of proteins. Traditional methods exploiting either sequence similarity or structural similarity can be unreliable and cannot be applied to proteins with novel folds or low homology with other proteins. Here, we present a machine-learning application that combines computed electrostatic, evolutionary, and pocket geometric information for high-performance prediction of catalytic residues. Input features consist of our structure-based theoretical microscopic anomalous titration curve shapes (THEMATICS) electrostatics data, enhanced with sequence-based phylogenetic information from INTREPID and topological pocket information from ConCavity. Our THEMATICS-based input features are augmented with an additional metric, the theoretical buffer range. With the integration of the three different types of input, each of which performs admirably on its own, significantly better performance is achieved than that of any of these methods by itself. This combined method achieves 86.7%, 92.5%, and 93.8% recall of annotated functional residues at 5, 8, and 10% false-positive rates, respectively.

A Tale of Two Isomerases: Compact versus Extended Active Sites in Ketosteroid Isomerase and Phosphoglucose Isomerase

Understanding the catalytic efficiency and specificity of enzymes is a fundamental question of major practical and conceptual importance in biochemistry. Although progress in biochemical and structural studies has enriched our knowledge of enzymes, the role in enzyme catalysis of residues that are not nearest neighbors of the reacting substrate molecule is largely unexplored experimentally. Here computational active site predictors, THEMATICS and POOL, were employed to identify functionally important residues that are not in direct contact with the reacting substrate molecule. These predictions then guided experiments to explore the active sites of two isomerases, Pseudomonas putida ketosteroid isomerase (KSI) and human phosphoglucose isomerase (PGI), as prototypes for very different types of predicted active sites. Both KSI and PGI are members of EC 5.3 and catalyze similar reactions, but they represent significantly different degrees of remote residue participation, as predicted by THEMATICS and POOL. For KSI, a compact active site of mostly first-shell residues is predicted, but for PGI, an extended active site in which residues in the first, second, and third layers around the reacting substrate are predicted. Predicted residues that have not been previously tested experimentally were investigated by site-directed mutagenesis and kinetic analysis. In human PGI, single-point mutations of the predicted second- and third-shell residues K362, H100, E495, D511, H396, and Q388 show significant decreases in catalytic activity relative to that of the wild type. The results of these experiments demonstrate that, as predicted, remote residues are very important in PGI catalysis but make only small contributions to catalysis in KSI.

Functional classification of protein 3D structures from predicted local interaction sites.

A new approach to the functional classification of protein 3D structures is described with application to some examples from structural genomics. This approach is based on functional site prediction with THEMATICS and POOL. THEMATICS employs calculated electrostatic potentials of the query structure. POOL is a machine learning method that utilizes THEMATICS features and has been shown to predict accurate, precise, highly localized interaction sites. Extension to the functional classification of structural genomics proteins is now described. Predicted functionally important residues are structurally aligned with those of proteins with previously characterized biochemical functions. A 3D structure match at the predicted local functional site then serves as a more reliable predictor of biochemical function than an overall structure match. Annotation is confirmed for a structural genomics protein with the ribulose phosphate binding barrel (RPBB) fold. A putative glucoamylase from Bacteroides fragilis (PDB ID 3eu8) is shown to be in fact probably not a glucoamylase. Finally a structural genomics protein from Streptomyces coelicolor annotated as an enoyl-CoA hydratase (PDB ID 3g64) is shown to be misannotated. Its predicted active site does not match the well-characterized enoyl-CoA hydratases of similar structure but rather bears closer resemblance to those of a dehalogenase with similar fold.

Solving nucleic acid structures by molecular replacement: examples from group II intron studies

Strategies for phasing nucleic acid structures by molecular replacement, using both experimental and de novo designed models, are discussed.

Crystal structure of a metal-dependent phosphoesterase (YP_910028.1) from Bifidobacterium adolescentis: Computational prediction and experimental validation of phosphoesterase activity.

The crystal structures of an unliganded and adenosine 5′‐monophosphate (AMP) bound, metal‐dependent phosphoesterase (YP_910028.1) from Bifidobacterium adolescentis are reported at 2.4 and 1.94 Å, respectively. Functional characterization of this enzyme was guided by computational analysis and then confirmed by experiment. The structure consists of a polymerase and histidinol phosphatase (PHP, Pfam: PF02811) domain with a second domain (residues 105‐178) inserted in the middle of the PHP sequence. The insert domain functions in binding AMP, but the precise function and substrate specificity of this domain are unknown. Initial bioinformatics analyses yielded multiple potential functional leads, with most of them suggesting DNA polymerase or DNA replication activity. Phylogenetic analysis indicated a potential DNA polymerase function that was somewhat supported by global structural comparisons identifying the closest structural match to the alpha subunit of DNA polymerase III. However, several other functional predictions, including phosphoesterase, could not be excluded. Theoretical microscopic anomalous titration curve shapes, a computational method for the prediction of active sites from protein 3D structures, identified potential reactive residues in YP_910028.1. Further analysis of the predicted active site and local comparison with its closest structure matches strongly suggested phosphoesterase activity, which was confirmed experimentally. Primer extension assays on both normal and mismatched DNA show neither extension nor degradation and provide evidence that YP_910028.1 has neither DNA polymerase activity nor DNA‐proofreading activity. These results suggest that many of the sequence neighbors previously annotated as having DNA polymerase activity may actually be misannotated. Proteins 2011.

Progress and Current Challenges in Modeling Large RNAs.

Recent breakthroughs in next-generation sequencing technologies have led to the discovery of several classes of non-coding RNAs (ncRNAs). It is now apparent that RNA molecules are not only just carriers of genetic information but also key players in many cellular processes. While there has been a rapid increase in the number of ncRNA sequences deposited in various databases over the past decade, the biological functions of these ncRNAs are largely not well understood. Similar to proteins, RNA molecules carry out a function by forming specific three-dimensional structures. Understanding the function of a particular RNA therefore requires a detailed knowledge of its structure. However, determining experimental structures of RNA is extremely challenging. In fact, RNA-only structures represent just 1% of the total structures deposited in the PDB. Thus, computational methods that predict three-dimensional RNA structures are in high demand. Computational models can provide valuable insights into structure-function relationships in ncRNAs and can aid in the development of functional hypotheses and experimental designs. In recent years, a set of diverse RNA structure prediction tools have become available, which differ in computational time, input data and accuracy. This review discusses the recent progress and challenges in RNA structure prediction methods.

Prediction of distal residue participation in enzyme catalysis

A scoring method for the prediction of catalytically important residues in enzyme structures is presented and used to examine the participation of distal residues in enzyme catalysis. Scores are based on the Partial Order Optimum Likelihood (POOL) machine learning method, using computed electrostatic properties, surface geometric features, and information obtained from the phylogenetic tree as input features. Predictions of distal residue participation in catalysis are compared with experimental kinetics data from the literature on variants of the featured enzymes; some additional kinetics measurements are reported for variants of Pseudomonas putida nitrile hydratase (ppNH) and for Escherichia coli alkaline phosphatase (AP). The multilayer active sites of P. putida nitrile hydratase and of human phosphoglucose isomerase are predicted by the POOL log ZP scores, as is the single‐layer active site of P. putida ketosteroid isomerase. The log ZP score cutoff utilized here results in over‐prediction of distal residue involvement in E. coli alkaline phosphatase. While fewer experimental data points are available for P. putida mandelate racemase and for human carbonic anhydrase II, the POOL log ZP scores properly predict the previously reported participation of distal residues.