[Paper]
This repository contains all of the code and data included in the article "Data-Driven Design of Protein-Like Single-Chain Polymer Nanoparticles" published on Chemrxiv Sep 13th, 2022.
The data directory includes one .csv files that contain data regarding physical attributes of SCNPs captured using dynamic light scattering (DLS), small angle x-ray scattering (SAXS), and chemical descriptor calculations. This is the data utilized in the manuscript to train evidential regression models for SCNP property prediction.
This code only requires a standard computer with enough RAM to support in-memory operations. However, as multiple machine learning pipelines occur here, we recommend running the code on systems with sufficient RAM and processing capacity:
RAM: 16+ GB
CPU: 4+ cores, 3.0+ GHz / core
Runtimes can be further improved through the usage of dedicated GPUs for model training.
The code was developed and tested only on Windows 10.
numpy
scikit-learn
pandas
keras
evidential-deep-learning
tensorflow
The easiest way to utilize the code is to download and run the following file:
scnp_demo.py
This demo file will train a simple evidential neural network for the immediate prediction of SCNP porod exponents. The code sources the data from the our data folder and will run 50 epochs of training for a predefined evidential neural network on 80% of the total dataset. Then, after training it will predict the scaled porod exponents of the held out data, passing the results to a local results.csv file.
This demo may be further expanded as we continue to update this reposiory.
The specific application, data, and machine learning models are described in: Data Driven Design of Single Chain Polymer Nanoparticles by ^Upadhya, R.; ^Tamasi, M.J.; Di Mare, E.; Murthy, N.S.; *Gormley, A.J (^ denotes equal contributions, * denotes corresponding authors), Chemrxiv, 2022.
If you need help, have suggestions, identify issues, or have corrections, please send your comments to Prof. Adam Gormley at adam.gormley@rutgers.edu
This project is covered under the Apache 2.0 License.