LedPred: an R/bioconductor package to predict regulatory sequences using support vector machines.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Supervised classification using Support Vector Machines (SVMs) is effective for predicting cis-regulatory modules (CRMs).
• Existing tools lack integration of heterogeneous data types like position-specific scoring matrices, ChIP-seq data, and conservation scores for CRM prediction.
• There is a need for a comprehensive tool to leverage diverse biological data for accurate regulatory sequence identification.

Purpose of the Study:

• To develop LedPred, a flexible SVM-based workflow for predicting novel regulatory sequences.
• To integrate heterogeneous data sources for enhanced CRM prediction accuracy.
• To provide an accessible R/Bioconductor package and online server for widespread use.

Main Methods:

• Implemented a Support Vector Machine (SVM) workflow (LedPred) in R/Bioconductor.
• Integrated diverse feature types including position-specific scoring matrices, ChIP-seq data, and conservation scores.
• Developed an associated online server to facilitate access and usability.

Main Results:

• LedPred successfully predicts new regulatory sequences by leveraging known CRM annotations and diverse feature types.
• The tool demonstrates superior performance in predicting regulatory sequences in Drosophila and mouse datasets compared to existing SVM-based software.
• The integrated approach effectively utilizes heterogeneous data for improved prediction accuracy.

Conclusions:

• LedPred offers a powerful and flexible solution for cis-regulatory module prediction.
• The integration of heterogeneous data significantly enhances the accuracy of regulatory sequence identification.
• LedPred provides a valuable resource for genomic research, accessible via R/Bioconductor and an online server.