Articles written in Journal of Biosciences
Volume 41 Issue 4 December 2016 pp 743-750 ARTICLE
Pluripotency is a unique property of stem cells that allows them to differentiate into all types of adult cells or maintainthe self-renewal property. PluriPred predicts whether a protein is involved in pluripotency from primary proteinsequence using manually curated pluripotent proteins as training datasets. Machine learning techniques (MLTs) suchas Support Vector Machine (SVM), Naïve Base (NB), Random Forest (RF), and sequence alignment techniqueBLAST were used in our study. The combination of SVM and PSI-BLAST was our proposed best model, whichobtained a sensitivity of 77.40%, specificity of 79.72%, accuracy of 79.2%, and area under the ROC curve was 0.82using 5-fold cross-validation. Furthermore, PluriPred gives the confidence of the prediction from training dataset’sSVM score distribution and p-value from BLAST. We validated our proposed model with the other existing highthroughputstudies using blind/independent datasets. Using PluriPred, 233 novel core and 323 novel extended corepluripotent proteins from mouse proteome, and 167 novel core and 385 extended core pluripotent proteins fromhuman proteome, were predicted with high confidence. The Web application of PluriPred is available from bicresources.jcbose.ac.in/ssaha4/pluripred/. Many pluripotent genes/proteins take part in protein-protein networks associatedwith stem cell, cancer, and developmental biology, and we believe that PluriPred will help in these research.
Volume 44 Issue 4 September 2019 Article ID 0104 Review
Protein–protein interactions (PPIs) are important for the study of protein functions and pathways involved in differentbiological processes, as well as for understanding the cause and progression of diseases. Several high-throughput experimentaltechniques have been employed for the identification of PPIs in a few model organisms, but still, there is a huge gapin identifying all possible binary PPIs in an organism. Therefore, PPI prediction using machine-learning algorithms hasbeen used in conjunction with experimental methods for discovery of novel protein interactions. The two most popularsupervised machine-learning techniques used in the prediction of PPIs are support vector machines and random forestclassifiers. Bayesian-probabilistic inference has also been used but mainly for the scoring of high-throughput PPI datasetconfidence measures. Recently, deep-learning algorithms have been used for sequence-based prediction of PPIs. Severalclustering methods such as hierarchical and k-means are useful as unsupervised machine-learning algorithms for theprediction of interacting protein pairs without explicit data labelling. In summary, machine-learning techniques have beenwidely used for the prediction of PPIs thus allowing experimental researchers to study cellular PPI networks.
Volume 44 | Issue 5
Click here for Editorial Note on CAP Mode