JPLE: Joint Protein-Ligand Embedding

JPLE is a Python package for modeling protein-nucleic acid interactions using peptide k-mer and nucleic acid k-mer representations. It supports training, protein queries, and nucleic acid queries.

Installation

JPLE can be installed as a Python package via pip. It is recommended to use a dedicated Python environment (e.g., Conda).

# Create a Conda environment (optional)
conda create -n jple python=3.8 -y
conda activate jple

# Install JPLE from the local directory
pip install .

Dependencies are installed automatically, including:

• logomaker
• matplotlib
• numpy
• pandas
• pyhmmer
• scipy

After installation, the jple command will be available:

jple --help

Data

Raw and processed data for JPLE is under data. You can regenerate the processed data using the Jupyter notebook notebooks/preprocess_data.ipynb.

Raw data

Located in data/raw:

• hmm_acc.txt: Accessions of the 10 selected Pfam profile HMMs.
• rnacompete_metadata_eupri.tsv: Metadata for the 420 RBP constructs in EuPRI.
• RRMs_3D_hmm_extended.hmm: Custom RRM_1 profile HMM.
• zscore_eupri.tsv: Binding profiles of the 420 RBP constructs in EuPRI (requires unzipping).

Processed data

Located in data/processed:

• zscore_train.tsv: Binding profiles of the 348 RBP constructs used for training (requires unzipping).
• seq_train.fasta: Sequences of the 348 RBP constructs used for training.
• param_train.npz: Saved JPLE model parameters (requires generation by training).
• domain_rbp.hmm: Profile HMMs of the 10 selected domains plus the custom RRM_1 profile.

Usage

JPLE can be run via a Command-Line Interface (CLI) or a Python API.

It supports three modes:

1. Training
2. Protein query
3. Nucleic acid query

Example input and output files are provided in test.

Training

Train JPLE on protein sequences and their binding profiles.

Parameters are saved to a .npz file.

CLI

jple \
--mode train \
--param data/processed/param_train.npz \
--fasta data/processed/seq_train.fasta \
--hmm data/processed/domain_rbp.hmm \
--zscore data/processed/zscore_train.tsv

API

from jple import run_jple

protein_dict_dict = run_jple.main(
    mode='train',
    param_path='data/processed/param_train.npz',
    fasta_path='data/processed/seq_train.fasta',
    hmm_path='data/processed/domain_rbp.hmm',
    y_path='data/processed/zscore_train.tsv'
)

The returned protein_dict_dict is a dictionary keyed by protein name (FASTA header).

Each inner dictionary contains:

1. protein_seq: Protein sequence.
2. domain_df: Domains boundaries (if present).

Protein query

Predict binding profiles for protein sequences.

CLI

jple \
--mode predict_protein \
--param data/processed/param_train.npz \
--fasta test/data/seq_test.fasta \
--hmm data/processed/domain_rbp.hmm \
--output test/output_protein

API

from jple import run_jple

protein_dict_dict = run_jple.main(
    mode='predict_protein',
    param_path='data/processed/param_train.npz',
    fasta_path='test/data/seq_test.fasta',
    hmm_path='data/processed/domain_rbp.hmm',
)

Outputs

Results are saved to the --output directory (CLI) or returned in protein_dict_dict (API).

Output	CLI file	API key
Protein sequence	–	protein_seq
Domain boundaries	domain.tsv	domain_df
Binding profiles	zscore.tsv	zscore_df
JPLE e-dist	dist.txt	dist
Nearest neighbors	neighbor.tsv	neighbor_df
PWM	pwm.txt	pwm
IUPAC motif	iupac.txt	iupac
Sequence logo	logo.png	–

Notes:

• Domain boundaries are only reported if the protein contains selected domains.
• Everything downstream are produced only if the protein contains an RRM or KH domain.

Nucleic acid query

Predict residue importance profiles given binding profiles.

CLI

jple \
--mode predict_na \
--param data/processed/param_train.npz \
--zscore test/data/zscore_test.tsv \
--output test/output_na

API

from jple import run_jple

protein_dict_dict = run_jple.main(
    mode='predict_na',
    param_path='data/processed/param_train.npz',
    y_path='test/data/zscore_test.tsv'
)

Outputs

Results are saved to the --output directory (CLI) or returned in protein_dict_dict (API).

Output	CLI file	API key
Residue importance	importance.tsv	importance_df
JPLE e-dist	dist.txt	dist

License

This project is licensed under the BSD 3-Clause License. See LICENSE for details.

References

If you use JPLE in publications, please cite:

Sasse, A., Ray, D., Laverty, K.U. et al. A resource of RNA-binding protein motifs across eukaryotes reveals evolutionary dynamics and gene-regulatory function.

Nat Biotechnol (2025). https://doi.org/10.1038/s41587-025-02733-6