Social networks to biological networks: systems biology of Mycobacterium tuberculosis

Machine Learning for Prediction of Drug Targets in Microbe Associated Cardiovascular Diseases by Incorporating Host-pathogen Interaction Network Parameters

Host-pathogen interactions play a crucial role in invasion, infection, and induction of immune response in humans. In this work, four machine learning algorithms, namely Logistic regression, K-nearest neighbor, Support Vector Machine, and Random Forest were implemented for the classification of drug targets. The algorithms were trained using 3400 hosts and 3800 pathogen drug and non-drug target proteins as learning instances. For each protein, 68 pathogen and 73 host features were computed that included sequence, structure, biological and host-pathogen network centrality characteristics. The Random Forest classifier model achieved the best accuracy after 10-fold cross-validation. 99 % accuracy was achieved with a ROC-AUC score of 0.99±0.01 for both pathogen and host training sets. The Eigenvector Centrality of host-pathogen interactions and host-host interactions was the top feature in performing classification of pathogen and host targets respectively. Other features important for classification were the presence of catalytic and binding sites, low instability/aliphatic index, and cellular location. The Random Forest classifier was then used for prediction of drug targets involved in Microbe Associated Cardiovascular Diseases. 331 host and 743 pathogen proteins were predicted as drug targets by the random forest model and can be validated experimentally for therapeutic intervention in Microbe Associated Cardiovascular Diseases.

Novel MicroRNA Biomarkers for Colorectal Cancer Early Diagnosis and 5-Fluorouracil Chemotherapy Resistance but Not Prognosis: A Study from Databases to AI-Assisted Verifications

Colorectal cancer (CRC) is one of the major causes of cancer death worldwide. In general, early diagnosis for CRC and individual therapy have led to better survival for the cancer patients. Accumulating studies concerning biomarkers have provided positive evidence to improve cancer early diagnosis and better therapy. It is, however, still necessary to further investigate the precise biomarkers for cancer early diagnosis and precision therapy and predicting prognosis. In this study, AI-assisted systems with bioinformatics algorithm integrated with microarray and RNA sequencing (RNA-seq) gene expression (GE) data has been approached to predict microRNA (miRNA) biomarkers for early diagnosis of CRC based on the miRNA-messenger RNA (mRNA) interaction network. The relationships between the predicted miRNA biomarkers and other biological components were further analyzed on biological networks. Bayesian meta-analysis of diagnostic test was utilized to verify the diagnostic value of the miRNA candidate biomarkers and the combined multiple biomarkers. Biological function analysis was performed to detect the relationship of candidate miRNA biomarkers and identified biomarkers in pathways. Text mining was used to analyze the relationships of predicted miRNAs and their target genes with 5-fluorouracil (5-FU). Survival analyses were conducted to evaluate the prognostic values of these miRNAs in CRC. According to the number of miRNAs single regulated mRNAs (NSR) and the number of their regulated transcription factor gene percentage (TFP) on the miRNA-mRNA network, there were 12 promising miRNA biomarkers were selected. There were five potential candidate miRNAs (miRNA-186-5p, miRNA-10b-5, miRNA-30e-5p, miRNA-21 and miRNA-30e) were confirmed as CRC diagnostic biomarkers, and two of them (miRNA-21 and miRNA-30e) were previously reported. Furthermore, the combinations of the five candidate miRNAs biomarkers showed better prediction accuracy for CRC early diagnosis than the single miRNA biomarkers. miRNA-10b-5p and miRNA-30e-5p were associated with the 5-FU therapy resistance by targeting the related genes. These miRNAs biomarkers were not statistically associated with CRC prognosis.

Data Intensive Genome Level Analysis for Identifying Novel, Non-Toxic Drug Targets for Multi Drug Resistant Mycobacterium tuberculosis

We report the construction of a novel Systems Biology based virtual drug discovery model for the prediction of non-toxic metabolic targets in Mycobacterium tuberculosis (Mtb). This is based on a data-intensive genome level analysis and the principle of conservation of the evolutionarily important genes. In the 1623 sequenced Mtb strains, 890 metabolic genes identified through a systems approach in Mtb were evaluated for non-synonymous mutations. The 33 genes showed none or one variation in the entire 1623 strains, including 1084 Russian MDR strains. These invariant targets were further evaluated for their experimental and in silico essentiality as well as availability of their crystal structure in Protein Data Bank (PDB). Along with this, targets for the common existing antibiotics and the new Tb drug candidates were also screened for their variation across 1623 strains of Mtb for understanding the drug resistance. We propose that the reduced set of these reported targets could be a more effective starting point for medicinal chemists in generating new chemical leads. This approach has the potential of fueling the dried up Tuberculosis (Tb) drug discovery pipeline.

Mining Functional Modules in Heterogeneous Biological Networks Using Multiplex PageRank Approach

Identification of functional modules/sub-networks in large-scale biological networks is one of the important research challenges in current bioinformatics and systems biology. Approaches have been developed to identify functional modules in single-class biological networks; however, methods for systematically and interactively mining multiple classes of heterogeneous biological networks are lacking. In this paper, we present a novel algorithm (called mPageRank) that utilizes the Multiplex PageRank approach to mine functional modules from two classes of biological networks. We demonstrate the capabilities of our approach by successfully mining functional biological modules through integrating expression-based gene-gene association networks and protein-protein interaction networks. We first compared the performance of our method with that of other methods using simulated data. We then applied our method to identify the cell division cycle related functional module and plant signaling defense-related functional module in the model plant Arabidopsis thaliana. Our results demonstrated that the mPageRank method is effective for mining sub-networks in both expression-based gene-gene association networks and protein-protein interaction networks, and has the potential to be adapted for the discovery of functional modules/sub-networks in other heterogeneous biological networks. The mPageRank executable program, source code, the datasets and results of the presented two case studies are publicly and freely available at http://plantgrn.noble.org/MPageRank/.

Systems level mapping of metabolic complexity in Mycobacterium tuberculosis to identify high-value drug targets

BACKGROUND

The effectiveness of current therapeutic regimens for Mycobacterium tuberculosis (Mtb) is diminished by the need for prolonged therapy and the rise of drug resistant/tolerant strains. This global health threat, despite decades of basic research and a wealth of legacy knowledge, is due to a lack of systems level understanding that can innovate the process of fast acting and high efficacy drug discovery.

METHODS

The enhanced functional annotations of the Mtb genome, which were previously obtained through a crowd sourcing approach was used to reconstruct the metabolic network of Mtb in a bottom up manner. We represent this information by developing a novel Systems Biology Spindle Map of Metabolism (SBSM) and comprehend its static and dynamic structure using various computational approaches based on simulation and design.

RESULTS

The reconstructed metabolism of Mtb encompasses 961 metabolites, involved in 1152 reactions catalyzed by 890 protein coding genes, organized into 50 pathways. By accounting for static and dynamic analysis of SBSM in Mtb we identified various critical proteins required for the growth and survival of bacteria. Further, we assessed the potential of these proteins as putative drug targets that are fast acting and less toxic. Further, we formulate a novel concept of metabolic persister genes (MPGs) and compared our predictions with published in vitro and in vivo experimental evidence. Through such analyses, we report for the first time that de novo biosynthesis of NAD may give rise to bacterial persistence in Mtb under conditions of metabolic stress induced by conventional anti-tuberculosis therapy. We propose such MPG's as potential combination of drug targets for existing antibiotics that can improve their efficacy and efficiency for drug tolerant bacteria.

CONCLUSION

The systems level framework formulated by us to identify potential non-toxic drug targets and strategies to circumvent the issue of bacterial persistence can substantially aid in the process of TB drug discovery and translational research.