Deep Molecular Representation Learning via Fusing Physical and Chemical Information

Authors: Shuwen Yang, Ziyao Li, Guojie Song, Lingsheng Cai

NeurIPS 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We evaluated Phys Chem on quantum mechanics (QM7, QM8 and QM9) and physical chemistry (LIPOP, FREESOLV and ESOL) datasets, as well as drug effectiveness datasets of the notorious SARS-CoV-2. [...] We split all datasets into 8:1:1 as training, validation and test sets.
Researcher Affiliation Academia 1Key Laboratory of Machine Perception and Intelligence (MOE), Peking University, Beijing, China 2Center for Data Science, Peking University, Beijing, China
Pseudocode No No pseudocode or clearly labeled algorithm block found in the paper.
Open Source Code No The paper does not provide any explicit statement or link indicating the availability of open-source code for the described methodology.
Open Datasets Yes The datasets are publicly available at http://moleculenet.ai/datasets-1. The datasets are available at https://opendata.ncats.nih.gov/covid19/ (CC BY 4.0 license) and are continuously extended.
Dataset Splits Yes We split all datasets into 8:1:1 as training, validation and test sets.
Hardware Specification No No specific hardware details (like GPU models, CPU types, or memory) used for running experiments are mentioned in the paper.
Software Dependencies Yes We use the 2020.03.1.0 version of the RDKit package at http://www.rdkit.org/.
Experiment Setup Yes Unless otherwise specified, we used L = 2 pairs of blocks for Phys Chem. In the initializer, we used a 2-layer GCN and a 2-layer LSTM. In each Phys Net block, we set df = 3, S = 4 and τ = 0.25. In each Chem Net block, we set the dimensions of atom and bond states as 128 and 64, correspondingly. In the representation readout block, we used T = 1 global attentive layers with 256-dimensional meta-atom states.