AI Paper Atlas

Research map / Graph learning / Graph self-supervised learning

Does GNN Pretraining Help Molecular Representation?

Ruoxi Sun, Hanjun Dai, Adams Wei Yu

NeurIPS 2022 · Graph learning › Graph self-supervised learning · approach: Molecular graph pretraining · 20 citations (OpenAlex)

Read the paper

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add self-supervised pretraining (node, context, motif, contrastive)GIN with rich features · MoleculeNet 5 datasets, balanced scaffoldaverage gain over no pretrain: 0.122 / 0.126 / 0.128 / -0.578No clear effect
add supervised ChEMBL pretraining after self-supervisedGIN with rich features · MoleculeNet, balanced scaffoldaverage gain over no pretrain: 0 -> 1.638Improved
switch from balanced scaffold to scaffold split (self-supervised pretraining) (ablation)GIN with rich features · MoleculeNetaverage gain over no pretrain: Node Prediction 0.122 -> 0.990Mixed
use basic features instead of rich features (ablation)no-pretrain GIN · MoleculeNet, balanced scaffoldaverage ROC-AUC: 78.0 -> 76.7Hurt
use basic features instead of rich features (ablation)no-pretrain GIN · MoleculeNet, scaffold splitaverage ROC-AUC: 71.8 -> 68.7Hurt
add context-prediction pretraining when only basic features are used (scaffold) (ablation)GIN basic features · MoleculeNet, scaffoldaverage gain over no pretrain: 1.792Improved
pretrain on SAVI (about 500x larger than ZINC15) (ablation)GIN pretraining · MoleculeNet, balanced scaffold and scaffoldaverage gain over no pretrain: Context Prediction 0.126 -> 0.546No clear effect
tune learning rate per setup and average over three splits instead of defaults (ablation)pretraining evaluation protocol · MoleculeNetpretraining benefitHurt
GraphSAGE backbone instead of GIN (ablation)graph pretraining · MoleculeNetaverage gain over no pretrainNo clear effect
7 GNN layers instead of 5 (ablation)node-prediction pretraining · MoleculeNetROC-AUCNo clear effect
use released GROVER pretrained weights instead of training from scratchGROVER base · BACEROC-AUC: 86.2 -> 85.1No clear effect

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy