sxLaep

• Fast inference — pre-trained XGBoost classifier (799 estimators, max depth 9) optimized for rapid enzyme prediction on large-scale protein datasets
• Rich feature extraction — combines Pse-AAC, CTD, windowed AAC, and sequence length for comprehensive sequence representation (258 features)
• Easy to use — dual Python API and CLI interface; install via pip and start predicting in minutes with the bundled pre-trained model
• Bioinformatics ready — built-in FASTA parsing, batch processing, and parallelized feature extraction (joblib) for production workflows
• Docker support — pre-configured image with Python 3.11 and all dependencies for isolated execution

Pipeline

Dependencies

From PyPI

pip install sxlaep

From Source

git clone https://github.com/labxscut/sxLaep
cd sxLaep
pip install -e .

Docker

docker pull labxscut/sxlaep:python3.11
docker run --rm -it -v $(pwd):/workspace labxscut/sxlaep:python3.11 bash

Command Line (shorthand)

sxlaep --input sequences.fasta --output predictions.csv

Python API

from pathlib import Path
import sxlaep
from sxlaep.model import load_model, predict_sequences, predict_fasta

# Load the bundled pre-trained model (native UBJSON format)
model_path = Path(sxlaep.__file__).parent / "enzyme_xgb_model.ubj"
model = load_model(model_path)

# Single sequence prediction
df = predict_sequences(model, ["MKVLW..."])
# df['pred_label']          -> 0 (Non-Enzyme) or 1 (Enzyme)
# df['enzyme_probability']  -> float (0.0 ~ 1.0)

# FASTA file prediction
result = predict_fasta(
    model_path=model_path,
    fasta_path="sequences.fasta",
    output_csv="predictions.csv",
)
print(result.head())

Prediction Example

# CLI (shorthand, bundled model)
sxlaep --input data/query.fasta --output results/predictions.csv

# CLI (subcommand, custom model)
sxlaep predict \
  --model results/sxlaep_training/enzyme_xgb_model.ubj \
  --fasta data/query.fasta \
  --output results/query_predictions.csv

# Python API
from sxlaep.model import predict_fasta

df = predict_fasta(
    model_path="results/sxlaep_training/enzyme_xgb_model.ubj",
    fasta_path="data/query.fasta",
    output_csv="results/query_predictions.csv",
)
print(df[["sequence_id", "pred_label", "enzyme_probability"]].head())

Feature Extraction Parameters

Output Format (predict_fasta CSV)

Notes

• The pre-trained model (enzyme_xgb_model.ubj) is bundled with the package in native XGBoost UBJSON format for cross-version compatibility.
• Sequences containing non-standard amino acids are automatically sanitized (non-AA20 characters are stripped).
• Short sequences (< 10 amino acids) may yield less reliable predictions.
• Feature extraction supports parallel processing via n_jobs — set to -1 when processing large datasets.
• The bundled model uses 799 estimators with max depth 9; feature parameters are fixed (lag=10, weight=0.05, segments=3) and cannot be changed in shorthand CLI mode.

Citations

Please cite sxLaep if you use it in your research:

labxscut. sxLaep: Rapid Enzyme/Non-Enzyme Prediction using Sequence Features and XGBoost. (2026).

Additional references:

1. Chou KC. Pseudo amino acid composition and its applications in bioinformatics, proteomics and genome biology. Curr Protein Pept Sci. 2010;11(6):573-95.
2. Dubchak I, Muchnik I, Holbrook SR, Kim SH. Prediction of protein folding class using global description of amino acid sequence. Proc Natl Acad Sci USA. 1995;92(19):8700-4.
3. Chen T, Guestrin C. XGBoost: A scalable tree boosting system. KDD 2016. p.785-94.