AI Paper Atlas

Research map / Probabilistic and scientific modeling / Molecules and proteins

HEIST: A Graph Foundation Model for Spatial Transcriptomics and Proteomics Data

Hiren Madhu, João Felipe Rocha, Tinglin Huang, Siddharth Viswanath, Smita Krishnaswamy, Rex Ying

ICLR 2026 · Probabilistic and scientific modeling › Molecules and proteins · approach: Protein representation learning

Read the paperCheck your idea against 137,217 papers

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
fine-tune the HEIST decoder instead of zero-shot useHEIST · gene imputation, SkinPearson correlation: 0.350 -> 0.807Improved
add hierarchical modeling with cross-level message passing (full HEIST vs removing cross message passing) (ablation)HEIST · Charville outcomeAUC-ROC: 0.625 -> 0.861Improved
add pre-training (ablation)HEIST · Charville outcomeAUC-ROC: 0.500 -> 0.861Improved
add spatial cell graph level (hierarchy) (ablation)HEIST · SEA-AD cell classificationF1: 0.179 -> 0.995Improved
add gene co-expression level (hierarchy) (ablation)HEIST · Skin imputationPearson correlation: 0.173 -> 0.807Improved
add masked autoencoder objective (ablation)HEIST · SEA-AD cell classificationF1: 0.162 -> 0.995Improved
add contrastive objective (ablation)HEIST · SEA-AD cell classificationF1: 0.966 -> 0.995Improved
add positional encodings (ablation)HEIST · SEA-AD cell classificationF1: 0.220 -> 0.995Improved
add orthogonal regularization (ablation)HEIST · Skin imputationPearson correlation: 0.646 -> 0.807Improved
use sinusoidal positional encodings instead of Laplacian or random-walk PEs (ablation)HEIST · ablation of positional encoding typesdownstream performanceImproved
use adaptive instead of constant co-expression threshold tau (ablation)HEIST · ablation of graph thresholddownstream performanceImproved
pass raw features through MLP before adding (ablation)HEIST · gene imputationimputation performanceHurt

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