PARSEbp: Pairwise Agreement-based RNA Scoring with Emphasis on Base Pairings

by Sumit Tarafder and Debswapna Bhattacharya

[bioRxiv] [pdf]

Codebase for our Pairwise Agreement-based RNA Scoring with Emphasis on Base Pairings (PARSEbp).

Installation

pip install PARSEbp

Or

git clone https://github.com/Bhattacharya-Lab/PARSEbp.git
cd PARSEbp
pip install .

Typical installation time should take less than a minute in a 64-bit Linux system.

Usage

Instructions for running PARSEbp:

# Import
from PARSEbp import parsebp

# Initialize
p = parsebp()

# Set target sequence to "" for sequence-agnostic scoring
p.set_target_sequnece("")

# Load a directory containing RNA 3D structures (.pdb files)
p.load_pdbs("Inputs/")

# Compute scores
score = p.score()

# Save the results
score.save("score.txt")

Additional functionality

# Set scoring mode (default is 1)
p.set_mode(1)

# Set target sequence (only the models that exactly match the target sequence will be scored)
seq = "GGACACGAGUAACUCGUCUAUCUGCUGCAGGCUGCUUACGGUUUCGUCCGUGUUGCAGCCGAUCAUCAGAACAUCUAGGUUUCGUCCGGGUGUUACCGAAAGGUCAGAUGGAGAGCCUUGUCCC"
p.set_target_sequnece(seq)

# Set the number of threads for parallel computations (default is 50)
p.set_parallel_threads(50)

# Get score of a specific model
score.getScore("decoy_1.pdb")

# Get top-1 ranked model(s)
score.top1()

# Get top-N ranked decoys
score.topN(10)

Given a directory containing RNA 3D structures "Inputs" as input, PARSEbp predicts the quality score of each structure in the directory and saves the output in "Score.txt".

Score calculation for a typical RNA (~100 nucleotides) with ~200 3D structures takes ~30 seconds.

Detailed Instructions

Follow the provided notebook for detailed explanation of the installation, scoring and score analysis.

Datasets

• CASP16 3D decoy structures (all submitted predictions) are downloaded from here.
• Ground truth scores for benchmarking PARSEbp is downloaded from here