Modeling

Modeling#

The modeling module within the FAMP pipeline is designed for de novo RNA structure prediction. Two external tools are employed:

  • RNAfold for secondary structure (2D) prediction

  • Rosetta (FARFAR2) for tertiary structure (3D) prediction

All modeling-related functions are encapsulated in the Modeling class. This module forms the first step in the FAMP pipeline and enables generation of structural input models for FRET-based analysis.

Initialization#

To start modeling, initialize the class by providing:

  • working_dir: Output directory for generated files

  • file_path_sequence: Path to the input RNA sequence in FASTA format

  • modeling_parameter: Python dictionary defining the modeling parameters (see the parameter section)

import famp
import os

current_dir = os.getcwd()
modeling = famp.modeling.Modeling(
    working_dir=f"{current_dir}",
    file_path_sequence=f"{current_dir}/input_data/RNA_Hairpin.fasta",
    modeling_parameter=rosetta_parameter
)

The input FASTA file should contain a single RNA sequence in the following format:

>Example_Sequence
caauauuuauuaauaucuuccggauauuaauaaauauug

Secondary Structure Prediction (2D)#

The function predict_2d_structure() performs secondary structure prediction using RNAfold. The RNA sequence is read from the input FASTA file, and the result is stored as a dot-bracket notation file.

  • Output directory: secondary_prediction/

  • Output file: dot_bracket.secstruct

This file can be edited manually after prediction to define or modify secondary structures.

modeling.predict_2d_structure()

Tertiary Structure Prediction (3D)#

3D structure prediction is performed using Rosetta’s FARFAR2 module via the predict_3d_structure() function. It requires the RNA secondary structure file (from dot_bracket.secstruct)

  • Output directory: rosetta_results

  • Output file: silent_out.out

modeling.predict_3d_structure("dot_bracket.secstruct")

Extracting PDB Structures#

Once Rosetta modeling is complete, the extract_pdb() function is used to convert the generated structure files into PDB format.

  • The models are ranked by Rosetta’s “res4” score.

  • The function returns the top n structures, where n is defined by number_of_structures.

modeling.extract_pdb(number_of_structures=5)

The resulting .pdb files are stored in the output directory and can be used for subsequent MD simulations or in silico dye labeling.

Contents