TYLER

TYLER predicts cyclin-dependent proteins (CDPs) from amino acid sequences to identify regulators of cell cycle progression relevant to cancer research and biomedical engineering.

Key Features:

• Sequence-based motif identification: Uses information theory to compute selectively enriched motifs associated with CDPs and builds a motif-based model.
• Dual-model strategy: Applies a motif-based model when enriched motifs are present and computes sequence-derived features with machine learning models for proteins lacking those motifs.
• Weight optimization and validation: Optimizes weights between the motif-based and machine learning models and validates performance using 5-fold cross-validation on a training dataset and independent test datasets.
• High-throughput performance: Demonstrates efficient runtime suitable for large-scale proteomic analyses.
• Empirical validation: Shows improved performance relative to existing methods, with predictions on the human proteome yielding novel CDP hypotheses supported by Gene Ontology (GO) analysis and some experimental verification.

Scientific Applications:

• Cancer research: Identification of CDPs to elucidate mechanisms of cell cycle regulation and uncontrolled proliferation in cancer.
• Biomedical engineering: Informing development of therapeutic interventions targeting cell cycle dysregulation.
• Proteome-wide CDP discovery and annotation: Enabling large-scale prediction of CDPs (including human proteome analyses) and downstream functional inference via Gene Ontology (GO) analysis.

Methodology:

Uses information theory to detect motifs selectively enriched in CDPs and builds a motif-based model; computes sequence-derived features for proteins lacking those motifs and trains machine learning models; optimizes model weights and evaluates performance with 5-fold cross-validation on a training dataset and independent test datasets.