ML-2DUV

This repository contains the the demo code for the secondary structure recognition using 2DUV spectra, as described in the paper "Machine learning recognition of protein secondary structures based on two-dimensional spectroscopic descriptors" (PNAS, 2022, DOI: 10.1073/pnas.2202713119). The whole data set were deposited on BaiDu Drive and Google Drive (see below), due to the large size.

created: 2022/04/28 @Zhang Qian

updated: 2022/05/01 @Hao Ren

Download dataset

We provide two datasets with different sizes: a shrinked set of 1.4 GB for test and demo use, and the whole dataset of ~30 GB. Both datasets can be accessed via clound drives.

We recommend downloading datasets of small size which just 1.4 GB: BaiDu Drive(extract code:PNAS) , Google Drive.

You can also download a full dataset in size of 30 GB: DCAIKU.

All Spectral data was simulated using method from our another repository (2duv_tutorial).

Data structure

Take 1.4G dataset as an example:

• original
  • original_dataset.npz
    • twoduv
    • la
    • cd
    • labels
  • original_transfer_dataset.npz
    • ...
• homologous
  • homologous_dataset.npz
    • ...
  • homologous_transfer_dataset.npz
    • ...
• nonhomologous
  • nonhomologous_dataset.npz
    • ...
  • nonhomologous_transfer_dataset.npz
    • ...

Prerequisite

Python 3.X
Tensorflow>=2.4.0
keras-tuner>=1.0.2
scikit-learn>=0.22
scikit-image>=0.16
numpy
pandas

Download code

git clone https://github.com/MTSD-UPC/ML-2DUV.git
cd ML-2DUV