Summary of Investigating Graph Neural Networks and Classical Feature-extraction Techniques in Activity-cliff and Molecular Property Prediction, by Markus Dablander
Investigating Graph Neural Networks and Classical Feature-Extraction Techniques in Activity-Cliff and Molecular Property Prediction
by Markus Dablander
First submitted to arxiv on: 20 Nov 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Biomolecules (q-bio.BM); Machine Learning (stat.ML)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary Molecular featurisation is a crucial process in machine learning and computational drug discovery. Recently, message-passing graph neural networks (GNNs) have emerged as a promising method for learning differentiable features from molecular graphs. While GNNs show great potential, it’s essential to investigate whether they can indeed outcompete classical molecular featurisation methods like extended-connectivity fingerprints (ECFPs) and physicochemical-descriptor vectors (PDVs). This study systematically explores and develops both classical and graph-based molecular featurisation methods for molecular property prediction, including quantitative structure-activity relationship (QSAR) prediction and activity-cliff (AC) prediction. The performance of PDVs, ECFPs, and GINs is compared for QSAR and AC-prediction using a rigorous computational study. Additionally, this paper introduces a novel twin neural network model for AC-prediction and proposes Sort & Slice, a simple substructure-pooling technique that outperforms hash-based folding at molecular property prediction. Finally, two ideas for future research are outlined: a graph-based self-supervised learning strategy to make classical molecular featurisations trainable, and trainable substructure-pooling via differentiable self-attention. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This study looks at how we can take information about molecules and turn it into numbers that computers can understand. This is important for developing new medicines and other products. The researchers looked at a few different ways to do this, including using special types of networks called graph neural networks (GNNs). They wanted to see if these GNNs are better than older methods for getting information about molecules. The study compared how well the different methods worked at predicting things like what happens when we mix certain chemicals together. The researchers also came up with a new way to do this and showed that it works well. Finally, they gave some ideas for how we can make these methods even better in the future. |
Keywords
» Artificial intelligence » Machine learning » Neural network » Self attention » Self supervised
