CSI-LSTM
This method predicts protein secondary structure from NMR chemical shifts using an LSTM model.
Citation: J Biomol NMR 2021, 75(10-12), 393-400.
Protein Science Software
Deep Learning Methods for Protein Science
This website presents software packages developed by Prof. Maili Liu's research group at APM, CAS. Our team develops machine learning and deep learning methods for protein science and NMR-related applications.
The original online services have been suspended. The software remains available through GitHub, and the links below point to the corresponding repositories and citations.
Method Library
Available Projects
Each method is listed with a short description, a GitHub repository link, and its citation.
OPPE
This method predicts protein order parameters from NMR structure ensembles using random forest.
Citation: J Biomol NMR 2024, 78, 87-94.
S-OPPE
An OPPE variant that improves performance through self-training.
Citation: ChemRxiv preprint, DOI 10.26434/chemrxiv-2025-grw5q.
SOPPCL
Sequence-based order parameter prediction with contrastive learning.
Citation: ChemRxiv preprint, DOI 10.26434/chemrxiv-2025-201mq.
RH-Unet
This method corrects peak shapes in 1D NMR spectra sampled in an inhomogeneous field using a U-net model.
Citation: Anal Chem 2023, 95(45), 16567-16574.
SENNet
This method separates NMR signals of small and large molecules from 1D NMR spectra using a U-net model.
Citation: Comm Chem 2024, 7(1), 167.
SSF-LSTM
This method predicts protein B-factors using an LSTM network.
Citation: ChemRxiv preprint, DOI 10.26434/chemrxiv-2023-59cp5.