LTHREADER: prediction of ligand-receptor interactions using localized threading.

Area of Science:

• Computational biology
• Bioinformatics
• Structural biology

Background:

• Identifying ligand-receptor interactions is crucial for drug discovery and disease treatment.
• High-throughput experimental methods face challenges in detecting these interactions, necessitating computational approaches.

Purpose of the Study:

• To introduce LTHREADER, a novel threading algorithm for accurate prediction of ligand-receptor interactions.
• To improve the accuracy of sequence-structure alignments and integrate diverse scoring methods for interaction prediction.

Main Methods:

• LTHREADER utilizes secondary structure and solvent accessibility predictions, along with residue contact maps, to guide alignments.
• A decision tree classifier trained on experimental data integrates statistical potentials, energy functions, correlated mutations, and conserved residue pairs.
• Predicted interaction significance is assessed using scores from randomized binding surfaces.

Main Results:

• LTHREADER demonstrates a 20% improvement in alignment accuracy of interacting residues compared to the state-of-the-art RAPTOR.
• For the 4-helical long chain cytokine family, LTHREADER achieved 75% sensitivity and 86% specificity in predicting interactions.
• The method shows promise for the TNF-like cytokine family with 70% sensitivity and 55% specificity, predicting novel interactions.

Conclusions:

• LTHREADER offers a significant advancement in computational prediction of ligand-receptor interactions.
• The algorithm's accuracy and ability to predict novel interactions have implications for drug design and understanding disease mechanisms.
• Application to cytokines highlights its potential in studying cancer and inflammatory/autoimmune disorders.