MULTICOM

MULTICOM predicts protein tertiary structures using a multi-template combination approach and large-scale model quality assessment to improve sampling, ranking, and accuracy of structural models.

Key Features:

• Multi-Template Combination: Integrates complementary and alternative templates, alignments, and models to average out errors and capture diverse structural information.
• Large-Scale Model Quality Assessment (QA): Employs 14 different model QA methods and generates consensus rankings to identify high-quality models.
• Model Clustering: Organizes similar structural models into clusters to support selection and downstream refinement.
• Model Refinement by Averaging: Refines clustered models through averaging to enhance model accuracy and reliability.

Scientific Applications:

• CASP11 (2014) benchmarking: Ranked third out of 143 human and server predictors by total scores of first models for 78 protein domains and second by best-of-five model predictions for those domains.
• CASP8 (2008) benchmarking: Multi-level combination was implemented across five automated servers and one human predictor, consistently ranking among top predictors.
• Target coverage: Capable of predicting moderate- to high-resolution models for most template-based targets and low-resolution models for some template-free targets.

Methodology:

Integrates complementary and alternative templates, alignments, and models (multi-template combination); applies 14 model QA methods to produce consensus rankings; performs model clustering; and refines models by averaging.