Pulse.

a daily field guide to health research that matters

◆ Console

‹ Sat · 16 May 2026
Promising but preliminary

HESpotEx: a dual-stream deep learning framework for spot-level gene expression prediction from histological images

AI trained on routine tissue slides can predict which genes are active in tumors without expensive genetic sequencing, democratizing molecular diagnostics.

HESpotEx uses a graph attention autoencoder + CNN decoder architecture to predict spatial transcriptomic profiles of up to 5,457 genes from routine histology slides, without expensive spatial RNA sequencing. Validated on TCGA and multiple cancer tissue types including lymphoma, it could democratize spatial molecular characterization of tumors.

What the study was

Study design
Computational validation study on multiple human cancer/non-cancer WSI datasets including TCGA
Population
Multiple cancer cohorts from TCGA and in-house WSI datasets; lymphoma included
Category
Genomics/Precision Medicine
Maturity
Exploratory
Journal
Nat Comput Sci

Why it surfaced

Nature Computational Science publication enabling cost-free spatial transcriptomics from routine histology slides — could fundamentally lower barriers to molecular tumor profiling. Validated on TCGA-scale data. Alert suppressed (cap reached at 5).

A plain-language summary of published research — not medical advice. Talk to a clinician about your own care.