FARFAR2

FARFAR2 predicts three-dimensional RNA structures de novo using fragment assembly and full-atom refinement to enable modeling of RNA folding and interactions.

Key Features:

• Fragment assembly with full-atom refinement: FARFAR2 assembles RNA fragments into full-atom models that capture detailed nucleotide geometries.
• Systematic benchmarking: FARFAR2 was evaluated by revisiting 21 RNA-Puzzles in blind tests without experimental data or expert intervention.
• Performance in blind trials: FARFAR2 produced models more accurate than original submissions in 16 of 21 RNA-Puzzle cases.
• Scoring function and sampling limitations: Accuracy is constrained by conformational sampling for RNAs longer than 80 nucleotides and by remaining scoring-function shortcomings.
• Preregistered blind models: Preregistered models of adenovirus VA-I RNA and five riboswitch complexes achieved native-like folds with RMSD accuracies of 3–14 Å.
• Updated fragment libraries and helix modeling: The method uses updated fragment libraries and enhanced helix modeling capabilities to improve assembly.
• Rosetta fragment-assembly framework: FARFAR2 builds upon Rosetta's fragment assembly framework for sampling and scoring RNA conformations.

Scientific Applications:

• RNA structural modeling: Predicts 3D RNA structures to support analysis of RNA folding and structural dynamics.
• Functional and interaction inference: Provides structural models to inform hypotheses about RNA function and molecular interactions.
• Modeling of viral RNAs and riboswitches: Applied to adenovirus VA-I RNA and multiple riboswitch complexes as demonstrated by preregistered blind predictions.

Methodology:

FARFAR2 uses Rosetta's fragment-assembly framework with updated fragment libraries, improved helix modeling, and full-atom refinement, and was benchmarked via blind revisitations of 21 RNA-Puzzle targets.