Dynamics of domain coverage of the protein sequence universe

Gene Ontology Capsule GAN: an improved architecture for protein function prediction

Proteins are the core of all functions pertaining to living things. They consist of an extended amino acid chain folding into a three-dimensional shape that dictates their behavior. Currently, convolutional neural networks (CNNs) have been pivotal in predicting protein functions based on protein sequences. While it is a technology crucial to the niche, the computation cost and translational invariance associated with CNN make it impossible to detect spatial hierarchies between complex and simpler objects. Therefore, this research utilizes capsule networks to capture spatial information as opposed to CNNs. Since capsule networks focus on hierarchical links, they have a lot of potential for solving structural biology challenges. In comparison to the standard CNNs, our results exhibit an improvement in accuracy. Gene Ontology Capsule GAN (GOCAPGAN) achieved an F1 score of 82.6%, a precision score of 90.4% and recall score of 76.1%.

DeepFunc: A Deep Learning Framework for Accurate Prediction of Protein Functions from Protein Sequences and Interactions

Annotation of protein functions plays an important role in understanding life at the molecular level. High-throughput sequencing produces massive numbers of raw proteins sequences and only about 1% of them have been manually annotated with functions. Experimental annotations of functions are expensive, time-consuming and do not keep up with the rapid growth of the sequence numbers. This motivates the development of computational approaches that predict protein functions. A novel deep learning framework, DeepFunc, is proposed which accurately predicts protein functions from protein sequence- and network-derived information. More precisely, DeepFunc uses a long and sparse binary vector to encode information concerning domains, families, and motifs collected from the InterPro tool that is associated with the input protein sequence. This vector is processed with two neural layers to obtain a low-dimensional vector which is combined with topological information extracted from protein-protein interactions (PPIs) and functional linkages. The combined information is processed by a deep neural network that predicts protein functions. DeepFunc is empirically and comparatively tested on a benchmark testing dataset and the Critical Assessment of protein Function Annotation algorithms (CAFA) 3 dataset. The experimental results demonstrate that DeepFunc outperforms current methods on the testing dataset and that it secures the highest F_max  = 0.54 and AUC = 0.94 on the CAFA3 dataset.

Compositionally Biased Dark Matter in the Protein Universe

Compositionally biased regions (BRs) occur when a few amino-acid types are enriched in a protein segment. There are possibly BR types in the known protein universe that have not been characterized experimentally. The UniProt protein database has been surveyed for evidence of such compositionally ''dark matter''. A ''dark biased region'' (DBR) is defined as a biased region with low probability of being an individual structural domain or intrinsically disordered region. The bias annotation program fLPS is used to generate a list of >13 million BRs, which is then thoroughly filtered for structure and intrinsic disorder. About a third of BRs (31%) has both substantial intrinsic disorder and structure. After filtering, there are ≈0.9 million DBRs (≈7% of the original BRs in ≈1.4% of proteins). These DBRs are hugely enriched in eukaryotes and hugely depleted in bacteria. They tend to be more hydrophobic than other protein regions, but are made of less extreme combinations of hydrophobic/hydrophilic residues. Given varying assumptions, It has been estimated that how many DBRs there might be for the high bias levels examined (with p-values < 1 × 10^-06 ), deriving a reasonable range of 0.7-7.2% of proteins having such DBRs. Hypotheses are examined about what such DBRs might be, that is, that they are from un- or undersampled domain/region categories or are unappreciated categories somewhat like existing ones.

Exploring the dark foldable proteome by considering hydrophobic amino acids topology

The protein universe corresponds to the set of all proteins found in all organisms. A way to explore it is by taking into account the domain content of the proteins. However, some part of sequences and many entire sequences remain un-annotated despite a converging number of domain families. The un-annotated part of the protein universe is referred to as the dark proteome and remains poorly characterized. In this study, we quantify the amount of foldable domains within the dark proteome by using the hydrophobic cluster analysis methodology. These un-annotated foldable domains were grouped using a combination of remote homology searches and domain annotations, leading to define different levels of darkness. The dark foldable domains were analyzed to understand what make them different from domains stored in databases and thus difficult to annotate. The un-annotated domains of the dark proteome universe display specific features relative to database domains: shorter length, non-canonical content and particular topology in hydrophobic residues, higher propensity for disorder, and a higher energy. These features make them hard to relate to known families. Based on these observations, we emphasize that domain annotation methodologies can still be improved to fully apprehend and decipher the molecular evolution of the protein universe.

Cache Domains That are Homologous to, but Different from PAS Domains Comprise the Largest Superfamily of Extracellular Sensors in Prokaryotes

Cellular receptors usually contain a designated sensory domain that recognizes the signal. Per/Arnt/Sim (PAS) domains are ubiquitous sensors in thousands of species ranging from bacteria to humans. Although PAS domains were described as intracellular sensors, recent structural studies revealed PAS-like domains in extracytoplasmic regions in several transmembrane receptors. However, these structurally defined extracellular PAS-like domains do not match sequence-derived PAS domain models, and thus their distribution across the genomic landscape remains largely unknown. Here we show that structurally defined extracellular PAS-like domains belong to the Cache superfamily, which is homologous to, but distinct from the PAS superfamily. Our newly built computational models enabled identification of Cache domains in tens of thousands of signal transduction proteins including those from important pathogens and model organisms. Furthermore, we show that Cache domains comprise the dominant mode of extracellular sensing in prokaryotes.

Cell-surface receptors control multiple cellular functions and are attractive targets for drug design. These receptors often have dedicated extracellular domains that bind signaling molecules, such as hormones and nutrients. Computational identification of these ligand-binding domains in genomic sequences is a pre-requisite for their further experimental characterization. Using available three-dimensional structures of several bacterial cell-surface receptors, we built computational models that enabled identification of the Cache domain, as the most common extracellular sensor module in prokaryotes, including many important pathogens. We also demonstrated that the Cache domain is homologous to, but sufficiently different from the most common intracellular sensor module, the PAS domain. These findings provide a unified view on molecular principles of signal recognition by extra- and intracellular receptors.

The language of the protein universe

Proteins, the main cell machinery which play a major role in nearly every cellular process, have always been a central focus in biology. We live in the post-genomic era, and inferring information from massive data sets is a steadily growing universal challenge. The increasing availability of fully sequenced genomes can be regarded as the 'Rosetta Stone' of the protein universe, allowing the understanding of genomes and their evolution, just as the original Rosetta Stone allowed Champollion to decipher the ancient Egyptian hieroglyphics. In this review, we consider aspects of the protein domain architectures repertoire that are closely related to those of human languages and aim to provide some insights about the language of proteins.

Most partial domains in proteins are alignment and annotation artifacts

BACKGROUND

Protein domains are commonly used to assess the functional roles and evolutionary relationships of proteins and protein families. Here, we use the Pfam protein family database to examine a set of candidate partial domains. Pfam protein domains are often thought of as evolutionarily indivisible, structurally compact, units from which larger functional proteins are assembled; however, almost 4% of Pfam27 PfamA domains are shorter than 50% of their family model length, suggesting that more than half of the domain is missing at those locations. To better understand the structural nature of partial domains in proteins, we examined 30,961 partial domain regions from 136 domain families contained in a representative subset of PfamA domains (RefProtDom2 or RPD2).

RESULTS

We characterized three types of apparent partial domains: split domains, bounded partials, and unbounded partials. We find that bounded partial domains are over-represented in eukaryotes and in lower quality protein predictions, suggesting that they often result from inaccurate genome assemblies or gene models. We also find that a large percentage of unbounded partial domains produce long alignments, which suggests that their annotation as a partial is an alignment artifact; yet some can be found as partials in other sequence contexts.

CONCLUSIONS

Partial domains are largely the result of alignment and annotation artifacts and should be viewed with caution. The presence of partial domain annotations in proteins should raise the concern that the prediction of the protein's gene may be incomplete. In general, protein domains can be considered the structural building blocks of proteins.

CDvist: a webserver for identification and visualization of conserved domains in protein sequences

SUMMARY

Identification of domains in protein sequences allows their assigning to biological functions. Several webservers exist for identification of protein domains using similarity searches against various databases of protein domain models. However, none of them provides comprehensive domain coverage while allowing bulk querying and their visualization schemes can be improved. To address these issues, we developed CDvist (a comprehensive domain visualization tool), which combines the best available search algorithms and databases into a user-friendly framework. First, a given protein sequence is matched to domain models using high-specificity tools and only then unmatched segments are subjected to more sensitive algorithms resulting in a best possible comprehensive coverage. Bulk querying and rich visualization and download options provide improved functionality to domain architecture analysis.

AVAILABILITY AND IMPLEMENTATION

Freely available on the web at http://cdvist.utk.edu

CONTACT

oadebali@vols.utk.edu or ijouline@utk.edu.

Unraveling the Complexities of Life Sciences Data

The life sciences have entered into the realm of big data and data-enabled science, where data can either empower or overwhelm. These data bring the challenges of the 5 Vs of big data: volume, veracity, velocity, variety, and value. Both independently and through our involvement with DELSA Global (Data-Enabled Life Sciences Alliance, DELSAglobal.org), the Kolker Lab ( kolkerlab.org ) is creating partnerships that identify data challenges and solve community needs. We specialize in solutions to complex biological data challenges, as exemplified by the community resource of MOPED (Model Organism Protein Expression Database, MOPED.proteinspire.org ) and the analysis pipeline of SPIRE (Systematic Protein Investigative Research Environment, PROTEINSPIRE.org ). Our collaborative work extends into the computationally intensive tasks of analysis and visualization of millions of protein sequences through innovative implementations of sequence alignment algorithms and creation of the Protein Sequence Universe tool (PSU). Pushing into the future together with our collaborators, our lab is pursuing integration of multi-omics data and exploration of biological pathways, as well as assigning function to proteins and porting solutions to the cloud. Big data have come to the life sciences; discovering the knowledge in the data will bring breakthroughs and benefits.