A sequence alignment-independent method for protein classification

Convolutional neural networks (CNNs): concepts and applications in pharmacogenomics

Convolutional neural networks (CNNs) have been used to extract information from various datasets of different dimensions. This approach has led to accurate interpretations in several subfields of biological research, like pharmacogenomics, addressing issues previously faced by other computational methods. With the rising attention for personalized and precision medicine, scientists and clinicians have now turned to artificial intelligence systems to provide them with solutions for therapeutics development. CNNs have already provided valuable insights into biological data transformation. Due to the rise of interest in precision and personalized medicine, in this review, we have provided a brief overview of the possibilities of implementing CNNs as an effective tool for analyzing one-dimensional biological data, such as nucleotide and protein sequences, as well as small molecular data, e.g., simplified molecular-input line-entry specification, InChI, binary fingerprints, etc., to categorize the models based on their objective and also highlight various challenges. The review is organized into specific research domains that participate in pharmacogenomics for a more comprehensive understanding. Furthermore, the future intentions of deep learning are outlined.

Classification of G-protein coupled receptors based on a rich generation of convolutional neural network, N-gram transformation and multiple sequence alignments

Sequence classification is crucial in predicting the function of newly discovered sequences. In recent years, the prediction of the incremental large-scale and diversity of sequences has heavily relied on the involvement of machine-learning algorithms. To improve prediction accuracy, these algorithms must confront the key challenge of extracting valuable features. In this work, we propose a feature-enhanced protein classification approach, considering the rich generation of multiple sequence alignment algorithms, N-gram probabilistic language model and the deep learning technique. The essence behind the proposed method is that if each group of sequences can be represented by one feature sequence, composed of homologous sites, there should be less loss when the sequence is rebuilt, when a more relevant sequence is added to the group. On the basis of this consideration, the prediction becomes whether a query sequence belonging to a group of sequences can be transferred to calculate the probability that the new feature sequence evolves from the original one. The proposed work focuses on the hierarchical classification of G-protein Coupled Receptors (GPCRs), which begins by extracting the feature sequences from the multiple sequence alignment results of the GPCRs sub-subfamilies. The N-gram model is then applied to construct the input vectors. Finally, these vectors are imported into a convolutional neural network to make a prediction. The experimental results elucidate that the proposed method provides significant performance improvements. The classification error rate of the proposed method is reduced by at least 4.67% (family level I) and 5.75% (family Level II), in comparison with the current state-of-the-art methods. The implementation program of the proposed work is freely available at: https://github.com/alanFchina/CNN .

Suite of tools for statistical N-gram language modeling for pattern mining in whole genome sequences

Genome sequences contain a number of patterns that have biomedical significance. Repetitive sequences of various kinds are a primary component of most of the genomic sequence patterns. We extended the suffix-array based Biological Language Modeling Toolkit to compute n-gram frequencies as well as n-gram language-model based perplexity in windows over the whole genome sequence to find biologically relevant patterns. We present the suite of tools and their application for analysis on whole human genome sequence.

A singular value decomposition approach for improved taxonomic classification of biological sequences

Background

Singular value decomposition (SVD) is a powerful technique for information retrieval; it helps uncover relationships between elements that are not prima facie related. SVD was initially developed to reduce the time needed for information retrieval and analysis of very large data sets in the complex internet environment. Since information retrieval from large-scale genome and proteome data sets has a similar level of complexity, SVD-based methods could also facilitate data analysis in this research area.

Results

We found that SVD applied to amino acid sequences demonstrates relationships and provides a basis for producing clusters and cladograms, demonstrating evolutionary relatedness of species that correlates well with Linnaean taxonomy. The choice of a reasonable number of singular values is crucial for SVD-based studies. We found that fewer singular values are needed to produce biologically significant clusters when SVD is employed. Subsequently, we developed a method to determine the lowest number of singular values and fewest clusters needed to guarantee biological significance; this system was developed and validated by comparison with Linnaean taxonomic classification.

Conclusions

By using SVD, we can reduce uncertainty concerning the appropriate rank value necessary to perform accurate information retrieval analyses. In tests, clusters that we developed with SVD perfectly matched what was expected based on Linnaean taxonomy.

FUNCTION PREDICTION OF UNCHARACTERIZED PROTEINS

Function prediction of uncharacterized protein sequences generated by genome projects has emerged as an important focus for computational biology. We have categorized several approaches beyond traditional sequence similarity that utilize the overwhelmingly large amounts of available data for computational function prediction, including structure-, association (genomic context)-, interaction (cellular context)-, process (metabolic context)-, and proteomics-experiment-based methods. Because they incorporate structural and experimental data that is not used in sequence-based methods, they can provide additional accuracy and reliability to protein function prediction. Here, first we review the definition of protein function. Then the recent developments of these methods are introduced with special focus on the type of predictions that can be made. The need for further development of comprehensive systems biology techniques that can utilize the ever-increasing data presented by the genomics and proteomics communities is emphasized. For the readers' convenience, tables of useful online resources in each category are included. The role of computational scientists in the near future of biological research and the interplay between computational and experimental biology are also addressed.

The accurate prediction of protein family from amino acid sequence by measuring features of sequence fragments

The rapid advances in proteomic analyses coupled with the completion of multiple genomes have led to an increased demand for determining protein functions. The first step is classification or prediction into families. A method was developed for the prediction of protein family based only on protein sequence using support vector machine (SVM) models. In these models, the amino acids were classified into three categories (apolar, polar, and charged). Consecutive fragments ranging from one to five were annotated by amino acid type to define the protein features of each protein. SVM models were constructed based on the protein features of a training set of proteins and then examined with an independent set of proteins. The approach was tested for 20 protein families from the iProClass database of Protein Information Resources (PIR). For two-class SVM models, an average prediction accuracy of 0.9985 was achieved, while for multi-class SVM models an accuracy of 0.9941 was achieved. This study demonstrates that SVM based methods can accurately recognize and predict the protein family to which a sequence belongs based solely on its primary amino acid sequence.

Subfamily specific conservation profiles for proteins based on n-gram patterns

Background

A new algorithm has been developed for generating conservation profiles that reflect the evolutionary history of the subfamily associated with a query sequence. It is based on n-gram patterns (NP{n,m}) which are sets of n residues and m wildcards in windows of size n+m. The generation of conservation profiles is treated as a signal-to-noise problem where the signal is the count of n-gram patterns in target sequences that are similar to the query sequence and the noise is the count over all target sequences. The signal is differentiated from the noise by applying singular value decomposition to sets of target sequences rank ordered by similarity with respect to the query.

Results

The new algorithm was used to construct 4,248 profiles from 120 randomly selected Pfam-A families. These were compared to profiles generated from multiple alignments using the consensus approach. The two profiles were similar whenever the subfamily associated with the query sequence was well represented in the multiple alignment. It was possible to construct subfamily specific conservation profiles using the new algorithm for subfamilies with as few as five members. The speed of the new algorithm was comparable to the multiple alignment approach.

Conclusion

Subfamily specific conservation profiles can be generated by the new algorithm without aprioi knowledge of family relationships or domain architecture. This is useful when the subfamily contains multiple domains with different levels of representation in protein databases. It may also be applicable when the subfamily sample size is too small for the multiple alignment approach.

Potential implications of availability of short amino acid sequences in proteins: an old and new approach to protein decoding and design

Three-dimensional structure of a protein molecule is primarily determined by its amino acid sequence, and thus the elucidation of general rules embedded in amino acid sequences is of great importance in protein science and engineering. To extract valuable information from sequences, we propose an analytical method in which a protein sequence is considered to be constructed by serial superimpositions of short amino acid sequences of n amino acid sets, especially triplets (3-aa sets). Using the comprehensive nonredundant protein database, we first examined "availability" of all possible combinatorial sets of 8,000 triplet species. Availability score was mathematically defined as an indicator for the relative "preference" or "avoidance" for a given short constituent sequence to be used in protein chain. Availability scores of real proteins were clearly biased against those of randomly generated proteins. We found many triplet species that occurred in the database more than expected or less than expected. Such bias was extended to longer sets, and we found that some species of pentats (5-aa sets) that occurred reasonably frequently in the randomly generated protein population did not occur at all in any real proteins known today. Availability score was dependent on species, potentially serving as a phylogenetic indicator. Furthermore, we suggest possibilities of various biotechnological applications of characteristic short sequences such as human-specific and pathogen-specific short sequences obtained from availability analysis. Availability score was also dependent on secondary structures, potentially serving as a structural indicator. Availability analysis on triplets may be combined with a comprehensive data collection on the varphi and psi peptide-bond angles of the amino acid at the center of each triplet, i.e., a collection of Ramachandran plots for each triplet. These triplet characters, together with other physicochemical data, will provide us with basic information between protein sequence and structure, by which structure prediction and engineering may be greatly facilitated. Availability analysis may also be useful in identifying word processing units in amino acid sequences based on an analogy to natural languages. Together with other approaches, availability analysis will elucidate general rules hidden in the primary sequences and eventually contributes to rebuilding the paradigm of protein science.

The relationship between n-gram patterns and protein secondary structure

An n-gram pattern (NP{n,m}) in a protein sequence is a set of n residues and m wildcards in a window of size n+m. Each window of n+m amino acids is associated with a collection of NP{n,m} patterns based on the combinatorics of n+m objects taken m at a time. NP{n,m} patterns that are shared between sequences reflect evolutionary relationships. Recently the authors developed an alignment-independent protein classification algorithm based on shared NP{4,2} patterns that compared favorably to PSI-BLAST. Theoretically, NP{4,2} patterns should also reflect secondary structure propensity since they contain all possible n-grams for 1 < or = n < or = 4 and a window of 6 residues is wide enough to capture periodicities in the 2 < or = n < or = 5 range. This sparked interest in differentiating the information content in NP{4,2} patterns related to evolution from the content related to local propensity. The probability of alpha-, beta-, and coil components was determined for every NP{4,2} pattern over all the chains in the Protein Data Bank (PDB). An algorithm exclusively based on the Z-values of these distributions was developed, which accurately predicted 71-76% of alpha-helical segments and 62-67% of beta-sheets in rigorous jackknife tests. This provided evidence for the strong correlation between NP{4,2} patterns and secondary structure. By grouping PDB chains into subsets with increasing levels of sequence identity, it was also possible to separate the evolutionary and local propensity contributions to the classification process. The results showed that information derived from evolutionary relationships was more important for beta-sheet prediction than alpha-helix prediction.

Recruitment of rare 3-grams at functional sites: is this a mechanism for increasing enzyme specificity?

Background

A wealth of unannotated and functionally unknown protein sequences has accumulated in recent years with rapid progresses in sequence genomics, giving rise to ever increasing demands for developing methods to efficiently assess functional sites. Sequence and structure conservations have traditionally been the major criteria adopted in various algorithms to identify functional sites. Here, we focus on the distributions of the 20³ different types of 3-grams (or triplets of sequentially contiguous amino acid) in the entire space of sequences accumulated to date in the UniProt database, and focus in particular on the rare 3-grams distinguished by their high entropy-based information content.

Results

Comparison of the UniProt distributions with those observed near/at the active sites on a non-redundant dataset of 59 enzyme/ligand complexes shows that the active sites preferentially recruit 3-grams distinguished by their low frequency in the UniProt. Three cases, Src kinase, hemoglobin, and tyrosyl-tRNA synthetase, are discussed in details to illustrate the biological significance of the results.

Conclusion

The results suggest that recruitment of rare 3-grams may be an efficient mechanism for increasing specificity at functional sites. Rareness/scarcity emerges as a feature that may assist in identifying key sites for proteins function, providing information complementary to that derived from sequence alignments. In addition it provides us (for the first time) with a means of identifying potentially functional sites from sequence information alone, when sequence conservation properties are not available.

Probabilistic annotation of protein sequences based on functional classifications

Background

One of the most evident achievements of bioinformatics is the development of methods that transfer biological knowledge from characterised proteins to uncharacterised sequences. This mode of protein function assignment is mostly based on the detection of sequence similarity and the premise that functional properties are conserved during evolution. Most automatic approaches developed to date rely on the identification of clusters of homologous proteins and the mapping of new proteins onto these clusters, which are expected to share functional characteristics.

Results

Here, we inverse the logic of this process, by considering the mapping of sequences directly to a functional classification instead of mapping functions to a sequence clustering. In this mode, the starting point is a database of labelled proteins according to a functional classification scheme, and the subsequent use of sequence similarity allows defining the membership of new proteins to these functional classes. In this framework, we define the Correspondence Indicators as measures of relationship between sequence and function and further formulate two Bayesian approaches to estimate the probability for a sequence of unknown function to belong to a functional class. This approach allows the parametrisation of different sequence search strategies and provides a direct measure of annotation error rates. We validate this approach with a database of enzymes labelled by their corresponding four-digit EC numbers and analyse specific cases.

Conclusion

The performance of this method is significantly higher than the simple strategy consisting in transferring the annotation from the highest scoring BLAST match and is expected to find applications in automated functional annotation pipelines.