Effects of windowing and zero-padding on Complex Resonant Recognition Model for protein sequence analysis

Harnessing the power of machine learning for crop improvement and sustainable production

Crop improvement and production domains encounter large amounts of expanding data with multi-layer complexity that forces researchers to use machine-learning approaches to establish predictive and informative models to understand the sophisticated mechanisms underlying these processes. All machine-learning approaches aim to fit models to target data; nevertheless, it should be noted that a wide range of specialized methods might initially appear confusing. The principal objective of this study is to offer researchers an explicit introduction to some of the essential machine-learning approaches and their applications, comprising the most modern and utilized methods that have gained widespread adoption in crop improvement or similar domains. This article explicitly explains how different machine-learning methods could be applied for given agricultural data, highlights newly emerging techniques for machine-learning users, and lays out technical strategies for agri/crop research practitioners and researchers.

BoT-Net: a lightweight bag of tricks-based neural network for efficient LncRNA–miRNA interaction prediction

Background and objective:

Interactions of long non-coding ribonucleic acids (lncRNAs) with micro-ribonucleic acids (miRNAs) play an essential role in gene regulation, cellular metabolic, and pathological processes. Existing purely sequence based computational approaches lack robustness and efficiency mainly due to the high length variability of lncRNA sequences. Hence, the prime focus of the current study is to find optimal length trade-offs between highly flexible length lncRNA sequences.

Method

The paper at hand performs in-depth exploration of diverse copy padding, sequence truncation approaches, and presents a novel idea of utilizing only subregions of lncRNA sequences to generate fixed-length lncRNA sequences. Furthermore, it presents a novel bag of tricks-based deep learning approach “Bot-Net” which leverages a single layer long-short-term memory network regularized through DropConnect to capture higher order residue dependencies, pooling to retain most salient features, normalization to prevent exploding and vanishing gradient issues, learning rate decay, and dropout to regularize precise neural network for lncRNA–miRNA interaction prediction.

Results

BoT-Net outperforms the state-of-the-art lncRNA–miRNA interaction prediction approach by 2%, 8%, and 4% in terms of accuracy, specificity, and matthews correlation coefficient. Furthermore, a case study analysis indicates that BoT-Net also outperforms state-of-the-art lncRNA–protein interaction predictor on a benchmark dataset by accuracy of 10%, sensitivity of 19%, specificity of 6%, precision of 14%, and matthews correlation coefficient of 26%.

Conclusion

In the benchmark lncRNA–miRNA interaction prediction dataset, the length of the lncRNA sequence varies from 213 residues to 22,743 residues and in the benchmark lncRNA–protein interaction prediction dataset, lncRNA sequences vary from 15 residues to 1504 residues. For such highly flexible length sequences, fixed length generation using copy padding introduces a significant level of bias which makes a large number of lncRNA sequences very much identical to each other and eventually derail classifier generalizeability. Empirical evaluation reveals that within 50 residues of only the starting region of long lncRNA sequences, a highly informative distribution for lncRNA–miRNA interaction prediction is contained, a crucial finding exploited by the proposed BoT-Net approach to optimize the lncRNA fixed length generation process.

Availability:

BoT-Net web server can be accessed at https://sds_genetic_analysis.opendfki.de/lncmiRNA/.

Graphic Abstract

Classification of Influenza Hemagglutinin Protein Sequences using Convolutional Neural Networks

The Influenza virus can be considered as one of the most severe viruses that can infect multiple species with often fatal consequences to the hosts. The Hemagglutinin (HA) gene of the virus can be a target for antiviral drug development realised through accurate identification of its sub-types and possible the targeted hosts. This paper focuses on accurately predicting if an Influenza type A virus can infect specific hosts, and more specifically, Human, Avian and Swine hosts, using only the protein sequence of the HA gene. In more detail, we propose encoding the protein sequences into numerical signals using the Hydrophobicity Index and subsequently utilising a Convolutional Neural Network-based predictive model. The Influenza HA protein sequences used in the proposed work are obtained from the Influenza Research Database (IRD). Specifically, complete and unique HA protein sequences were used for avian, human and swine hosts. The data obtained for this work was 17999 human-host proteins, 17667 avian-host proteins and 9278 swine-host proteins. Given this set of collected proteins, the proposed method yields as much as 10% higher accuracy for an individual class (namely, Avian) and 5% higher overall accuracy than in an earlier study. It is also observed that the accuracy for each class in this work is more balanced than what was presented in this earlier study. As the results show, the proposed model can distinguish HA protein sequences with high accuracy whenever the virus under investigation can infect Human, Avian or Swine hosts.

Prediction of Influenza A virus infections in humans using an Artificial Neural Network learning approach

The Influenza type A virus can be considered as one of the most severe viruses that can infect multiple species with often fatal consequences to the hosts. The Haemagglutinin (HA) gene of the virus has the potential to be a target for antiviral drug development realised through accurate identification of its sub-types and possible the targeted hosts. In this paper, to accurately predict if an Influenza type A virus has the capability to infect human hosts, by using only the HA gene, is therefore developed and tested. The predictive model follows three main steps; (i) decoding the protein sequences into numerical signals using EIIP amino acid scale, (ii) analysing these sequences by using Discrete Fourier Transform (DFT) and extracting DFT-based features, (iii) using a predictive model, based on Artificial Neural Networks and using the features generated by DFT. In this analysis, from the Influenza Research Database, 30724, 18236 and 8157 HA protein sequences were collected for Human, Avian and Swine respectively. Given this set of the proteins, the proposed method yielded 97.36% (± 0.04%), 97.26% (± 0.26%), 0.978 (± 0.004), 0.963 (± 0.005) and 0.945 (±0.005) for the training accuracy validation accuracy, precision, recall and Mathews Correlation Coefficient (MCC) respectively, based on a 10-fold cross-validation. The classification model generated by using one of the largest dataset, if not the largest, yields promising results that could lead to early detection of such species and help develop precautionary measurements for possible human infections.

Prediction of protein allergenicity based on signal-processing bioinformatics approach

Current bioinformatics tools accomplish high accuracies in classifying allergenic protein sequences with high homology and generally perform poorly with low homology protein sequences. Although some homologous regions explained Immunoglobulin E (IgE) cross-reactivity in groups of allergens, no universal molecular structure could be associated with allergenicity. In addition, studies have showed that cross-reactivity is not directly linked to the homology between protein sequences. Therefore, a new homology independent method needs to be developed to determine if a protein is an allergen or not. The aim of this study is therefore to differentiate sets of allergenic and non-allergenic proteins using a signal-processing based bioinformatics approach. In this paper, a new method was proposed for characterisation and classification of allergenic protein sequences. For this method hydrophobicity amino acid index was used to encode proteins to numerical sequences and Discrete Fourier Transform to extract features for each protein. Finally, a classifier was constructed based on Support Vector Machines. In order to demonstrate the applicability of the proposed method 857 allergen and 1000 non-allergen proteins were collected from UniProt online database. The results obtained from the proposed method yielded: MCC: 0.752 ± 0.007, Specificity: 0.912 ± 0.005, Sensitivity: 0.835 ± 0.008 and Total Accuracy: 87.65% ± 0.004.

Structural classification of protein sequences based on signal processing and support vector machines

The function of any protein depends directly on its secondary and tertiary structure. Proteins can fold into a three-dimensional shape, which is primarily depended on the arrangement of amino acids in the primary structure. In recent years, with the explosive sequencing of proteins, it is unfeasible to perform detailed experimental studies, as these methodologies are very expensive and time consuming. This leaves the structure of the majority of currently available protein sequences unknown. In this paper, a predictive model is therefore presented for the classification of protein sequence's secondary structures, namely alpha helix and beta sheet. The proteins used throughout this study were collected from the Structural Classification of Proteinsextended (SCOPe) database, which contains manually curated information from proteins with known structure. Two sets of proteins are used for all alpha and all beta protein sequences. The first set comprise of sequences with less than 40% identity, and the second set comprise of proteins with less than 95% identity. The analysis shows a strong connection between the amino acid indices used to convert protein sequences to numerical sequences and proteins' secondary structures. The total classification accuracy for the proposed classifier for the protein sequences with less than 40% identity for amino acid index BIOV880101 and BIOV880102 are 78.49% and 76.40%, respectively. The classification accuracy for sets of protein sequences with less than 95% identity for amino acid index BIOV880101 and BIOV880102 are 88.01% and 85.17%, respectively.

CISAPS: Complex Informational Spectrum for the Analysis of Protein Sequences

Complex informational spectrum analysis for protein sequences (CISAPS) and its web-based server are developed and presented. As recent studies show, only the use of the absolute spectrum in the analysis of protein sequences using the informational spectrum analysis is proven to be insufficient. Therefore, CISAPS is developed to consider and provide results in three forms including absolute, real, and imaginary spectrum. Biologically related features to the analysis of influenza A subtypes as presented as a case study in this study can also appear individually either in the real or imaginary spectrum. As the results presented, protein classes can present similarities or differences according to the features extracted from CISAPS web server. These associations are probable to be related with the protein feature that the specific amino acid index represents. In addition, various technical issues such as zero-padding and windowing that may affect the analysis are also addressed. CISAPS uses an expanded list of 611 unique amino acid indices where each one represents a different property to perform the analysis. This web-based server enables researchers with little knowledge of signal processing methods to apply and include complex informational spectrum analysis to their work.

Construction of protein distance matrix based on amino acid indices and Discrete Fourier Transform

Protein distance matrix is widely used in various protein sequence analyses, and mainly obtained by using pairwise sequence alignment scores or protein sequence homology, which fail to take into consideration of individual physical characteristics of protein sequences and amino acids, or a combination of these features. In this paper, a new method is therefore proposed for constructing protein distance matrix based on natural amino acid indices in combination with Discrete Fourier Transform (DFT). For the proposed method, protein distance matrices can be generated using any given set of amino acid indices, each one of which represents a unique biological feature of protein sequences. In this study, the results are based on the combination of 25 widely accepted amino acid indices, which produced the best results, according to the biological relationships between proteins. As a case study 26 Cluster of Differentiation 4 (CD4) protein sequences were used in order to construct a distance matrix based on the proposed method. The results show that the pairwise relationship between CD4 protein sequences remain the same in comparison with their pairwise percent identity. For another group of protein sequences the pairwise relationship between CD4 protein sequences dramatically changed with the proposed method in comparison to the pairwise percent identity. The proposed distance matrix has been shown to have a positive impact on these case studies and therefore is expected to be useful in several fields such as multiple protein sequence alignment and phylogenetic analysis, where an accurate distance matrix based on natural generalized protein properties plays an important role.