Clarion

Clarion predicts mRNA subcellular localization to identify multiple intracellular compartments of messenger RNAs and support analysis of spatial regulation of gene expression and gene regulatory networks.

Key Features:

• Multi-label multi-class prediction: Performs multi-label, multi-class prediction of mRNA subcellular localization to identify multiple simultaneous compartments.
• Weighted series transformation: Uses the weighted series method to transform multi-label problems into an analyzable format.
• Manually curated benchmark dataset: Trained and evaluated on a manually curated benchmark dataset of mRNAs.
• Independent benchmarking and performance metrics: Extensive independent tests report accuracies of 81.47% for chromatin, 91.29% for cytoplasm, 79.77% for cytosol, 92.10% for exosome, 89.15% for membrane, 83.74% for nucleolus, 80.74% for nucleoplasm, 79.23% for nucleus, and 84.74% for ribosome localizations.

Scientific Applications:

• Gene regulatory network analysis: Supports investigation of gene regulatory networks by providing compartment-specific mRNA localization predictions.
• Spatial regulation of gene activity: Enables studies of spatial regulation of gene expression by mapping mRNA distributions across intracellular compartments.
• Multi-localized mRNA studies: Facilitates identification and analysis of mRNAs localized to multiple cellular compartments simultaneously.

Methodology:

Implements a multi-label multi-class prediction framework using the weighted series method to transform multi-label problems and is trained and evaluated on a manually curated benchmark dataset with independent benchmarking tests reporting per-compartment accuracies.