AI for Interpretable Chemistry: Predicting Radical Mechanistic Pathways via Contrastive Learning
Authors: Mohammadamin Tavakoli, Pierre Baldi, Ann Marie Carlton, Yin Ting Chiu, Alexander Shmakov, David Van Vranken
NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We develop and train multiple deep-learning models using RMech DB, a public database of radical reactions, to establish the first benchmark for predicting radical reactions. Our results demonstrate the effectiveness of RMech RP in providing accurate and interpretable predictions of radical reactions, and its potential for various applications in atmospheric chemistry. |
| Researcher Affiliation | Academia | Mohammadamin Tavakoli Department of Computer Science University of California, Irvine mohamadt@uci.edu Yin Ting T.Chiu Department of Chemistry University of California, Irvine yintc@uci.edu Alexander Shmakov Department of Computer Science University of California, Irvine ashmakov@uci.edu Ann Marie Carlton Department of Chemistry University of California, Irvine agcarlto@uci.edu David Van Vranken Department of Chemistry University of California, Irvine david.vv@uci.edu Pierre Baldi Department of Computer Science University of California, Irvine pfbaldi@uci.edu |
| Pseudocode | No | The paper does not contain any pseudocode or clearly labeled algorithm blocks. |
| Open Source Code | No | Our radical reaction predictor is publicly available through an online interface at http://deeprxn.ics.uci.edu/rmechrp. |
| Open Datasets | Yes | We develop and train multiple deep-learning models using RMech DB, a public database of radical reactions, to establish the first benchmark for predicting radical reactions. |
| Dataset Splits | No | The paper mentions "standard train and test sets" and presents Table 1 with training and testing data sizes, but it does not explicitly provide details for a separate validation split or how one was handled for reproduction. |
| Hardware Specification | Yes | Finally, all the experiments are conducted using a single NVidia Titan X GPU. |
| Software Dependencies | No | The paper mentions using "RDKit [53]" and "Molecular Transformer [28]" but does not specify their version numbers or other key software dependencies with specific versions needed for full reproducibility. |
| Experiment Setup | Yes | Table 5: The parameters used for training the models for reactive sites identification. Table 6: The parameters used for training the models for the plausibility ranking. Table 7: The parameters used for training the Rxn-Hypergraph for the contrastive model. Both f and g have similar architectures that consist of three fully connected layers with a GELU activation function and a dropout with a rate of 0.5 applied to all layers. The dimensions of the layers are 128, 64, 1. |