PluriPred: AWeb server for predicting proteins involved in pluripotent network

Area of Science:

• * Stem cell biology
• * Bioinformatics
• * Computational biology

Background:

• * Pluripotency enables stem cells to differentiate into all adult cell types or self-renew.
• * Identifying proteins regulating pluripotency is crucial for stem cell research.
• * Existing methods may lack accuracy or efficiency in predicting pluripotent proteins.

Purpose of the Study:

• * To develop PluriPred, a computational tool for predicting protein involvement in pluripotency from primary sequences.
• * To utilize machine learning techniques and sequence alignment for accurate prediction.
• * To provide a user-friendly web application for researchers.

Main Methods:

• * Employed machine learning techniques including Support Vector Machine (SVM), Naïve Bayes (NB), and Random Forest (RF).
• * Utilized BLAST for sequence alignment.
• * Developed a hybrid model combining SVM and PSI-BLAST, trained on manually curated pluripotent proteins.

Main Results:

• * The SVM and PSI-BLAST model achieved 77.40% sensitivity, 79.72% specificity, and 79.2% accuracy with a 0.82 ROC AUC using 5-fold cross-validation.
• * PluriPred predicts novel core and extended core pluripotent proteins in mouse (233 core, 323 extended) and human (167 core, 385 extended) proteomes with high confidence.
• * Model performance was validated against independent datasets from high-throughput studies.

Conclusions:

• * PluriPred accurately predicts proteins involved in pluripotency, offering a valuable tool for stem cell research.
• * The identified novel pluripotent proteins can advance understanding of stem cell function, development, and related diseases like cancer.
• * The accessible web application facilitates broader research applications in bioinformatics and developmental biology.