DEGnext: classification of differentially expressed genes from RNA-seq data using a convolutional neural network with transfer learning.

BACKGROUND: A limitation of traditional differential expression analysis on small datasets involves the possibility of false positives and false negatives due to sample variation. Considering the recent advances in deep learning (DL) based models, we wanted to expand the state-of-the-art in disease biomarker prediction from RNA-seq data using DL. However, application of DL to RNA-seq data is challenging due to absence of appropriate labels and smaller sample size as compared to number of genes. Deep learning coupled with transfer learning can improve prediction performance on novel data by incorporating patterns learned from other related data. With the emergence of new disease datasets, biomarker prediction would be facilitated by having a generalized model that can transfer the knowledge of trained feature maps to the new dataset. To the best of our knowledge, there is no Convolutional Neural Network (CNN)-based model coupled with transfer learning to predict the significant upregulating (UR) and downregulating (DR) genes from both trained and untrained datasets. RESULTS: We implemented a CNN model, DEGnext, to predict UR and DR genes from gene expression data obtained from The Cancer Genome Atlas database. DEGnext uses biologically validated data along with logarithmic fold change values to classify differentially expressed genes (DEGs) as UR and DR genes. We applied transfer learning to our model to leverage the knowledge of trained feature maps to untrained cancer datasets. DEGnext's results were competitive (ROC scores between 88 and 99[Formula: see text]) with those of five traditional machine learning methods: Decision Tree, K-Nearest Neighbors, Random Forest, Support Vector Machine, and XGBoost. DEGnext was robust and effective in terms of transferring learned feature maps to facilitate classification of unseen datasets. Additionally, we validated that the predicted DEGs from DEGnext were mapped to significant Gene Ontology terms and pathways related to cancer. CONCLUSIONS: DEGnext can classify DEGs into UR and DR genes from RNA-seq cancer datasets with high performance. This type of analysis, using biologically relevant fine-tuning data, may aid in the exploration of potential biomarkers and can be adapted for other disease datasets.

X-Module: A novel fusion measure to associate co-expressed gene modules from condition-specific expression profiles.

A gene co-expression network (CEN) is of biological interest, since co-expressed genes share common functions and biological processes or pathways. Finding relationships among modules can reveal inter-modular preservation, and similarity in transcriptome, functional, and biological behaviors among modules of the same or two different datasets. There is no method which explores the one-to-one relationships and one-to-many relationships among modules extracted from control and disease samples based on both topological and semantic similarity using both microarray and RNA seq data. In this work, we propose a novel fusion measure to detect mapping between modules from two sets of co-expressed modules extracted from control and disease stages of Alzheimer's disease (AD) and Parkinson's disease (PD) datasets. Our measure considers both topological and biological information of a module and is an estimation of four parameters, namely, semantic similarity, eigengene correlation, degree difference, and the number of common genes. We analyze the consensus modules shared between both control and disease stages in terms of their association with diseases. We also validate the close associations between human and chimpanzee modules and compare with the state-ofthe- art method. Additionally, we propose two novel observations on the relationships between modules for further analysis.

Comparison of Methods for Differential Co-expression Analysis for Disease Biomarker Prediction.

In the recent past, a number of methods have been developed for analysis of biological data. Among these methods, gene co-expression networks have the ability to mine functionally related genes with similar co-expression patterns, because of which such networks have been most widely used. However, gene co-expression networks cannot identify genes, which undergo condition specific changes in their relationships with other genes. In contrast, differential co-expression analysis enables finding co-expressed genes exhibiting significant changes across disease conditions. In this paper, we present some significant outcomes of a comparative study of four co-expression network module detection techniques, namely, THD-Module Extractor, DiffCoEx, MODA, and WGCNA, which can perform differential co-expression analysis on both gene and miRNA expression data (microarray and RNA-seq) and discuss the applications to Alzheimer's disease and Parkinson's disease research. Our observations reveal that compared to other methods, THD-Module Extractor is the most effective in finding modules with higher functional relevance and biological significance.

THD-Tricluster: A robust triclustering technique and its application in condition specific change analysis in HIV-1 progression data.

Developing a cost-effective and robust triclustering algorithm that can identify triclusters of high biological significance in the gene-sample-time (GST) domain is a challenging task. Most existing triclustering algorithms can detect shifting and scaling patterns in isolation, they are not able to handle co-occurring shifting-and-scaling patterns. This paper makes an attempt to address this issue. It introduces a robust triclustering algorithm called THD-Tricluster to identify triclusters over the GST domain. In addition to applying over several benchmark datasets for its validation, the proposed THD-Tricluster algorithm was applied on HIV-1 progression data to identify disease-specific genes. THD-Tricluster could identify 38 most responsible genes for the deadly disease which includes GATA3, EGR1, JUN, ELF1, AGFG1, AGFG2, CX3CR1, CXCL12, CCR5, CCR2, and many others. The results are validated using GeneCard and other established results.

THD-Module Extractor: An Application for CEN Module Extraction and Interesting Gene Identification for Alzheimer's Disease.

There exist many tools and methods for construction of co-expression network from gene expression data and for extraction of densely connected gene modules. In this paper, a method is introduced to construct co-expression network and to extract co-expressed modules having high biological significance. The proposed method has been validated on several well known microarray datasets extracted from a diverse set of species, using statistical measures, such as p and q values. The modules obtained in these studies are found to be biologically significant based on Gene Ontology enrichment analysis, pathway analysis, and KEGG enrichment analysis. Further, the method was applied on an Alzheimer's disease dataset and some interesting genes are found, which have high semantic similarity among them, but are not significantly correlated in terms of expression similarity. Some of these interesting genes, such as MAPT, CASP2, and PSEN2, are linked with important aspects of Alzheimer's disease, such as dementia, increase cell death, and deposition of amyloid-beta proteins in Alzheimer's disease brains. The biological pathways associated with Alzheimer's disease, such as, Wnt signaling, Apoptosis, p53 signaling, and Notch signaling, incorporate these interesting genes. The proposed method is evaluated in regard to existing literature.