reMap

reMap predicts metabolic pathways from genomic sequence data to improve metabolic pathway inference across genomes of varying complexity and completion.

Key Features:

• Relabeling Framework: Transforms pathway data into groups characterized by statistical correlations among pathways, enabling multiple revisits of specific pathways across groups to improve sensitivity and accuracy.
• Alternating Feedback Process: Implements an alternating mechanism with a feed-forward phase that selects a minimal subset of pathway groups to label each genomic example and a feedback-backward phase that updates internal parameters to refine mappings.
• Multi-label Learning Algorithm Training: Produces a relabeled dataset that is used to train a multi-label learning algorithm for pathway prediction.

Scientific Applications:

• Pathway Prediction: Predicts metabolic pathways from genomic sequences using relabeled pathway-group representations and multi-label learning.
• Genome Complexity and Completion Handling: Applies to genomes with varying levels of complexity and completion by enabling iterative relabeling and mapping refinement.

Methodology:

reMap relabels pathway data into statistically correlated groups; applies an alternating feedback process with a feed-forward selection of minimal pathway-group subsets per genomic example and a feedback-backward parameter update; and produces a relabeled dataset used to train a multi-label learning algorithm.