AbDev is a comprehensive predictive model package designed for the analysis of 12 critical biophysical properties of monoclonal antibodies (mAbs). This tool combines a deep learning-based tool, DeepSP, and machine learning techniques to provide insights based on the variable regions sequences of mAbs.
To utilize AbDev effectively, follow the guidelines outlined below:
- Prepare a CSV File: Begin by preparing a CSV file named "Sequence_Info.csv" using the format provided in this guide as a reference. This file should contain the variable regions sequences of the mAbs you wish to analyze.
- Run the DeepSP Notebook: Use the "DeepSP.ipynb" notebook file to execute DeepSP, a deep learning-based tool developed by our group. DeepSP is designed to generate 30 spatial properties of mAbs based on their sequences.
- Upon completion, you will obtain a "SAPSCM.csv" file, which contains the spatial properties needed for further analysis.
- Run the AbDev Notebook: Next, execute the "AbDev.ipynb" notebook file. This step will process the features generated in the previous step and produce the "Prediction_Result.csv" file. This file includes predictions for 12 biophysical properties of the analyzed mAbs. Or you can run train.py to obtain the results directly.
- Please cite when using DeepSP in your research.
L. Kalejaye, I.E. Wu, T. Terry and P.K. Lai, DeepSP: Deep Learning-Based Spatial Properties to Predict Monoclonal Antibody Stability, Comput. Struct. Biotechnol. J., 23:2220–2229, 2024.
Read the DeepSP paper on CSBJ - Please cite when using AbDev in your research. I.E. Wu, L. Kalejaye and P.K. Lai, "Machine Learning Models for Predicting Monoclonal Antibody Biophysical Properties from Molecular Dynamics Simulations and Deep Learning-based Surface Descriptors", Mol. Pharm, 2024.
Read the AbDev paper on Mol. Pharm
