Deep learning and embeddings for problems of computational biology

The development of Next Generation Sequencing promotes Biology in the Big Data era. The ever-increasing gap between proteins with known sequences and those with a complete functional annotation requires computational methods for automatic structure and functional annotation. My research has been focusing on proteins and led so far to the development of three novel tools, DeepREx, E-SNPs&GO and ISPRED-SEQ, based on Machine and Deep Learning approaches. DeepREx computes the solvent exposure of residues in a protein chain. This problem is relevant for the definition of structural constraints regarding the possible folding of the protein. DeepREx exploits Long Short-Term Memory layers to capture residue-level interactions between positions distant in the sequence, achieving state-of-the-art performances. With DeepRex, I conducted a large-scale analysis investigating the relationship between solvent exposure of a residue and its probability to be pathogenic upon mutation. E-SNPs&GO predicts the pathogenicity of a Single Residue Variation. Variations occurring on a protein sequence can have different effects, possibly leading to the onset of diseases. E-SNPs&GO exploits protein embeddings generated by two novel Protein Language Models (PLMs), as well as a new way of representing functional information coming from the Gene Ontology. The method achieves state-of-the-art performances and is extremely time-efficient when compared to traditional approaches. ISPRED-SEQ predicts the presence of Protein-Protein Interaction sites in a protein sequence. Knowing how a protein interacts with other molecules is crucial for accurate functional characterization. ISPRED-SEQ exploits a convolutional layer to parse local context after embedding the protein sequence with two novel PLMs, greatly surpassing the current state-of-the-art. All methods are published in international journals and are available as user-friendly web servers. They have been developed keeping in mind standard guidelines for FAIRness (FAIR: Findable, Accessible, Interoperable, Reusable) and are integrated into the public collection of tools provided by ELIXIR, the European infrastructure for Bioinformatics.

New computational biology tools for the systematic analysis of the structure and expression of human genes

From the late 1980s, the automation of sequencing techniques and the computer spread gave rise to a flourishing number of new molecular structures and sequences and to proliferation of new databases in which to store them. Here are presented three computational approaches able to analyse the massive amount of publicly avalilable data in order to answer to important biological questions. The first strategy studies the incorrect assignment of the first AUG codon in a messenger RNA (mRNA), due to the incomplete determination of its 5' end sequence. An extension of the mRNA 5' coding region was identified in 477 in human loci, out of all human known mRNAs analysed, using an automated expressed sequence tag (EST)-based approach. Proof-of-concept confirmation was obtained by in vitro cloning and sequencing for GNB2L1, QARS and TDP2 and the consequences for the functional studies are discussed. The second approach analyses the codon bias, the phenomenon in which distinct synonymous codons are used with different frequencies, and, following integration with a gene expression profile, estimates the total number of codons present across all the expressed mRNAs (named here "codonome value") in a given biological condition. Systematic analyses across different pathological and normal human tissues and multiple species shows a surprisingly tight correlation between the codon bias and the codonome bias. The third approach is useful to studies the expression of human autism spectrum disorder (ASD) implicated genes. ASD implicated genes sharing microRNA response elements (MREs) for the same microRNA are co-expressed in brain samples from healthy and ASD affected individuals. The different expression of a recently identified long non coding RNA which have four MREs for the same microRNA could disrupt the equilibrium in this network, but further analyses and experiments are needed.

An integrated molecular biology and computational approach to CYP1A1 expression in human brain

The cytochromes P450 comprise a superfamily of heme-containing mono-oxygenases. These enzymes metabolize numerous xenobiotics, but also play a role in metabolism of endogenous compounds. The P450 1A1 enzyme generally metabolizes polycyclic aromatic hydrocarbons, and its expression can be induced by aryl hydrocarbon receptor (AhR) activation. CYP1A1 is an exception to the generality that the majority of CYPs demonstrate highest expression in liver; CYP1Al is present in numerous extrahepatic tissues, including brain. This P450 has been observed in two forms, wildtype (WT) and brain variant (BV), arising from alternatively spliced mRNA transcripts. The CYP1A1 BV mRNA presented an exon deletion and was detected in human brain but not liver tissue of the same individuals. ^ Quantitative PCR analyses were performed to determine CYP1A1 WT and BV transcript expression levels in normal, bipolar disorder or schizophrenic groups. In our samples, we show that CYP1A1 BV mRNA, when present, is found alongside the full-length form. Furthermore, we demonstrate a significant decrease in expression of CYP1A1 in patients with bipolar disorder or schizophrenia. The expression level was not influenced by post-mortem interval, tissue pH, age, tobacco use, or lifetime antipsychotic medication load. ^ There is no indication of increased brain CYP1A1 expression in normal smokers versus non-smokers in these samples. We observed slightly increased CYP1A1 expression only in bipolar and schizophrenic smokers versus non-smokers. This may be indicative of complex interactions between neuronal chemical environments and AhR-mediated CYP1A1 induction in brain. ^ Structural homology modeling demonstrated that P450 1A1 BV has several alterations to positions/orientations of substrate recognition site residues compared to the WT isoform. Automated substrate docking was employed to investigate the potential binding of neurological signaling molecules and neurotropic drugs, as well as to differentiate specificities of the two P450 1A1 isoforms. We consistently observed that the BV isoform produced energetically favorable substrate dockings in orientations not observed for the same substrate in the WT isoform. These results demonstrated that structural differences, namely an expanded substrate access channel and active site, confer greater capacity for unique compound docking positions suggesting a metabolic profile distinct from the wildtype form for these test compounds. ^

Functional and bioinformatics analyses reveal conservation of Cis-regulatory elements between sciaridae and Drosophilidae

The sciarid DNA puff C4 BhC4-1 gene is amplified and transcribed in salivary glands at the end of the larval stage. In transgenic Drosophila, the BhC4-1 promoter drives transcription in prepupal salivary glands and in the ring gland of late embryos. A bioinformatics analysis has identified 162 sequences similar to distinct regions of the BhC4-1 proximal promoter, which are predominantly located either in 5` or 3` regions or introns in the Drosophila melanogaster genome. A significant number of the identified sequences are found in the regulatory regions of Drosophila genes that are expressed in the salivary gland. Functional assays in Drosophila reveal that the BhC4-1 proximal promoter contains both a 129 bp (-186/-58) salivary gland enhancer and a 67 bp (-253/-187) ring gland enhancer that drive tissue specific patterns of developmentally regulated gene expression, irrespective of their orientation.

A checkpointing-enabled and resource-aware Java Virtual Machine for efficient and robust e-Science applications in grid environments

Object-oriented programming languages presently are the dominant paradigm of application development (e. g., Java,. NET). Lately, increasingly more Java applications have long (or very long) execution times and manipulate large amounts of data/information, gaining relevance in fields related with e-Science (with Grid and Cloud computing). Significant examples include Chemistry, Computational Biology and Bio-informatics, with many available Java-based APIs (e. g., Neobio). Often, when the execution of such an application is terminated abruptly because of a failure (regardless of the cause being a hardware of software fault, lack of available resources, etc.), all of its work already performed is simply lost, and when the application is later re-initiated, it has to restart all its work from scratch, wasting resources and time, while also being prone to another failure and may delay its completion with no deadline guarantees. Our proposed solution to address these issues is through incorporating mechanisms for checkpointing and migration in a JVM. These make applications more robust and flexible by being able to move to other nodes, without any intervention from the programmer. This article provides a solution to Java applications with long execution times, by extending a JVM (Jikes research virtual machine) with such mechanisms. Copyright (C) 2011 John Wiley & Sons, Ltd.

Towards data warehousing and mining of protein unfolding simulation data

OBJECTIVES: The prediction of protein structure and the precise understanding of protein folding and unfolding processes remains one of the greatest challenges in structural biology and bioinformatics. Computer simulations based on molecular dynamics (MD) are at the forefront of the effort to gain a deeper understanding of these complex processes. Currently, these MD simulations are usually on the order of tens of nanoseconds, generate a large amount of conformational data and are computationally expensive. More and more groups run such simulations and generate a myriad of data, which raises new challenges in managing and analyzing these data. Because the vast range of proteins researchers want to study and simulate, the computational effort needed to generate data, the large data volumes involved, and the different types of analyses scientists need to perform, it is desirable to provide a public repository allowing researchers to pool and share protein unfolding data. METHODS: To adequately organize, manage, and analyze the data generated by unfolding simulation studies, we designed a data warehouse system that is embedded in a grid environment to facilitate the seamless sharing of available computer resources and thus enable many groups to share complex molecular dynamics simulations on a more regular basis. RESULTS: To gain insight into the conformational fluctuations and stability of the monomeric forms of the amyloidogenic protein transthyretin (TTR), molecular dynamics unfolding simulations of the monomer of human TTR have been conducted. Trajectory data and meta-data of the wild-type (WT) protein and the highly amyloidogenic variant L55P-TTR represent the test case for the data warehouse. CONCLUSIONS: Web and grid services, especially pre-defined data mining services that can run on or 'near' the data repository of the data warehouse, are likely to play a pivotal role in the analysis of molecular dynamics unfolding data.

Relationship between global structural parameters and Enzyme Commission hierarchy: Implications for function prediction

Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP)

Identical sequence patterns in the ends of exons and introns of human protein-coding genes

Intron splicing is one of the most important steps involved in the maturation process of a pre-mRNA. Although the sequence profiles around the splice sites have been studied extensively, the levels of sequence identity between the exonic sequences preceding the donor sites and the intronic sequences preceding the acceptor sites has not been examined as thoroughly. In this study we investigated identity patterns between the last 15 nucleotides of the exonic sequence preceding the 5' splice site and the intronic sequence preceding the 3' splice site in a set of human protein-coding genes that do not exhibit intron retention. We found that almost 60% of consecutive exons and introns in human protein-coding genes share at least two identical nucleotides at their 3' ends and, on average, the sequence identity length is 2.47 nucleotides. Based on our findings we conclude that the 3' ends of exons and introns tend to have longer identical sequences within a gene than when being taken from different genes. Our results hold even if the pairs are non-consecutive in the transcription order. (C) 2012 Elsevier Ltd. All rights reserved.

Relationship between global structural parameters and Enzyme Commission hierarchy: Implications for function prediction

In protein databases there is a substantial number of proteins structurally determined but without function annotation. Understanding the relationship between function and structure can be useful to predict function on a large scale. We have analyzed the similarities in global physicochemical parameters for a set of enzymes which were classified according to the four Enzyme Commission (EC) hierarchical levels. Using relevance theory we introduced a distance between proteins in the space of physicochemical characteristics. This was done by minimizing a cost function of the metric tensor built to reflect the EC classification system. Using an unsupervised clustering method on a set of 1025 enzymes, we obtained no relevant clustering formation compatible with EC classification. The distance distributions between enzymes from the same EC group and from different EC groups were compared by histograms. Such analysis was also performed using sequence alignment similarity as a distance. Our results suggest that global structure parameters are not sufficient to segregate enzymes according to EC hierarchy. This indicates that features essential for function are rather local than global. Consequently, methods for predicting function based on global attributes should not obtain high accuracy in main EC classes prediction without relying on similarities between enzymes from training and validation datasets. Furthermore, these results are consistent with a substantial number of studies suggesting that function evolves fundamentally by recruitment, i.e., a same protein motif or fold can be used to perform different enzymatic functions and a few specific amino acids (AAs) are actually responsible for enzyme activity. These essential amino acids should belong to active sites and an effective method for predicting function should be able to recognize them. (C) 2012 Elsevier Ltd. All rights reserved.

Quantifying Reproducibility in Computational Biology: The Case of the Tuberculosis Drugome

How easy is it to reproduce the results found in a typical computational biology paper? Either through experience or intuition the reader will already know that the answer is with difficulty or not at all. In this paper we attempt to quantify this difficulty by reproducing a previously published paper for different classes of users (ranging from users with little expertise to domain experts) and suggest ways in which the situation might be improved. Quantification is achieved by estimating the time required to reproduce each of the steps in the method described in the original paper and make them part of an explicit workflow that reproduces the original results. Reproducing the method took several months of effort, and required using new versions and new software that posed challenges to reconstructing and validating the results. The quantification leads to “reproducibility maps” that reveal that novice researchers would only be able to reproduce a few of the steps in the method, and that only expert researchers with advance knowledge of the domain would be able to reproduce the method in its entirety. The workflow itself is published as an online resource together with supporting software and data. The paper concludes with a brief discussion of the complexities of requiring reproducibility in terms of cost versus benefit, and a desiderata with our observations and guidelines for improving reproducibility. This has implications not only in reproducing the work of others from published papers, but reproducing work from one’s own laboratory.

Bioinformatic analysis for exploring relationships between genes and gene products

To carry out their specific roles in the cell, genes and gene products often work together in groups, forming many relationships among themselves and with other molecules. Such relationships include physical protein-protein interaction relationships, regulatory relationships, metabolic relationships, genetic relationships, and much more. With advances in science and technology, some high throughput technologies have been developed to simultaneously detect tens of thousands of pairwise protein-protein interactions and protein-DNA interactions. However, the data generated by high throughput methods are prone to noise. Furthermore, the technology itself has its limitations, and cannot detect all kinds of relationships between genes and their products. Thus there is a pressing need to investigate all kinds of relationships and their roles in a living system using bioinformatic approaches, and is a central challenge in Computational Biology and Systems Biology. This dissertation focuses on exploring relationships between genes and gene products using bioinformatic approaches. Specifically, we consider problems related to regulatory relationships, protein-protein interactions, and semantic relationships between genes. A regulatory element is an important pattern or "signal", often located in the promoter of a gene, which is used in the process of turning a gene "on" or "off". Predicting regulatory elements is a key step in exploring the regulatory relationships between genes and gene products. In this dissertation, we consider the problem of improving the prediction of regulatory elements by using comparative genomics data. With regard to protein-protein interactions, we have developed bioinformatics techniques to estimate support for the data on these interactions. While protein-protein interactions and regulatory relationships can be detected by high throughput biological techniques, there is another type of relationship called semantic relationship that cannot be detected by a single technique, but can be inferred using multiple sources of biological data. The contributions of this thesis involved the development and application of a set of bioinformatic approaches that address the challenges mentioned above. These included (i) an EM-based algorithm that improves the prediction of regulatory elements using comparative genomics data, (ii) an approach for estimating the support of protein-protein interaction data, with application to functional annotation of genes, (iii) a novel method for inferring functional network of genes, and (iv) techniques for clustering genes using multi-source data.

REPRODUCTIVE BIOLOGY AND FUNCTIONAL MATURITY IN FEMALES OF AEGLA FRANCA (DECAPODA: ANOMURA: AEGLIDAE)

We describe the reproductive period. fecundity, and average size at the onset of functional maturity of female Aegla franca, the northernmost distributed aeglid species. The reproductive period is markedly seasonal and takes place front May (austral mid-autumn) to August (late winter). Ovigerous females appear quite abruptly in the population by May, and this condition is observed in all adult females sampled regardless of their size. The average size at the onset of functional maturity in females, at which 50% of the females sampled during the reproductive period were considered adults, was 12.75 mm CL. The smallest post-ovigerous female measured 12.06 mm carapace length (CL). Mean fecundity (+/- S.D.) from 41 females bearing early and intermediate eggs was 129.1 +/- 32.2 and corresponded to a mean female CL of 14.11 mm. The elliptical-shaped eggs exhibited significant increase in size along the development stages. The third pair of pleopods bore higher number of eggs than the others. Compiled information regarding the reproductive period reported for aeglids revealed all increase in the breeding period length with latitude. The reproductive period tends to be shorter in localities under larger rainfall variation and smaller temperature variability than in sites with opposite climate conditions. Eggs tend to be fewer in number and larger in size towards lower latitudes. We present an hypothesis that stream water velocity might act as a major selective pressure during the early life history of fluvial aeglids with direct effect on the reproductive pattern.

Complete Genome Sequence of Mycoplasma suis and Insights into Its Biology and Adaption to an Erythrocyte Niche

Mycoplasma suis, the causative agent of porcine infectious anemia, has never been cultured in vitro and mechanisms by which it causes disease are poorly understood. Thus, the objective herein was to use whole genome sequencing and analysis of M. suis to define pathogenicity mechanisms and biochemical pathways. M. suis was harvested from the blood of an experimentally infected pig. Following DNA extraction and construction of a paired end library, whole-genome sequencing was performed using GS-FLX (454) and Titanium chemistry. Reads on paired-end constructs were assembled using GS De Novo Assembler and gaps closed by primer walking; assembly was validated by PFGE. Glimmer and Manatee Annotation Engine were used to predict and annotate protein-coding sequences (CDS). The M. suis genome consists of a single, 742,431 bp chromosome with low G+C content of 31.1%. A total of 844 CDS, 3 single copies, unlinked rRNA genes and 32 tRNAs were identified. Gene homologies and GC skew graph show that M. suis has a typical Mollicutes oriC. The predicted metabolic pathway is concise, showing evidence of adaptation to blood environment. M. suis is a glycolytic species, obtaining energy through sugars fermentation and ATP-synthase. The pentose-phosphate pathway, metabolism of cofactors and vitamins, pyruvate dehydrogenase and NAD(+) kinase are missing. Thus, ribose, NADH, NADPH and coenzyme A are possibly essential for its growth. M. suis can generate purines from hypoxanthine, which is secreted by RBCs, and cytidine nucleotides from uracil. Toxins orthologs were not identified. We suggest that M. suis may cause disease by scavenging and competing for host nutrients, leading to decreased life-span of RBCs. In summary, genome analysis shows that M. suis is dependent on host cell metabolism and this characteristic is likely to be linked to its pathogenicity. The prediction of essential nutrients will aid the development of in vitro cultivation systems.