FRAGSION

FRAGSION generates protein fragment libraries without external databases to support de novo protein structure prediction, particularly template-free modeling (FM).

Key Features:

• Database-Free Approach: Eliminates dependence on external databases for fragment generation, avoiding limitations related to database size and availability.
• Efficiency and Speed: Generates a fragment library for a typical 300-residue protein in a few seconds of CPU time.
• Dynamic Fragment Generation: Produces fragments of arbitrary length, including fragments as long as the entire protein sequence.
• Noise Handling in Predicted Features: Incorporates mechanisms to handle noise in sequence-based predicted features such as secondary structure during fragment sampling.

Scientific Applications:

• De novo Structure Prediction (template-free modeling, FM): Supplies fragment libraries for de novo structure prediction workflows that rely on fragment-based sampling in FM approaches.
• Benchmarking and Method Comparison (CASP, ROSETTA): Enables benchmarking against methods like ROSETTA using datasets from CASP, demonstrating orders-of-magnitude speedups while maintaining comparable fragment quality.

Methodology:

FRAGSION uses an Input-Output Hidden Markov Model (IOHMM) framework for efficient sampling and generation of protein fragments and for managing noise in sequence-based predicted features.