Eldad David Shulman

This paper develops HEiST, a deep learning method to infer spatial transcriptomes from H&E slides, identifies two prognostic Spatiotypes in HNSC, and shows they predict treatment response better than existing biomarkers.

Path2Space predicts spatial gene expression and tumor microenvironment from histopathology slides, identifying new breast cancer subgroups and improving biomarker prediction.

SLIDE-EX predicts cell type-specific gene expression from breast cancer histopathology images, outperforming bulk RNA-seq deconvolution and slide-based models in chemotherapy response prediction.

Path2Omics predicts gene expression and methylation from histopathology slides across 23 cancer types, outperforming previous models and showing effectiveness in predicting patient outcomes.

This paper develops DECODEM to predict chemotherapy response by integrating deconvolved expression patterns of immune, stromal, and tumor cells in breast cancer, finding that cell-type-specific expression is predictive of response.

This paper develops a machine learning framework, EXPRESSO, to predict treatment response from tumor transcriptomics across multiple cancer types and therapies, showing improved performance over existing methods.

This paper introduces EXPAND, an interpretable AI model that predicts breast cancer subtypes and survival using nuclear features, achieving performance comparable to black-box models.

ecPath detects ecDNA in tumors using H&E images, achieving good performance in predicting ecDNA status and identifying poor prognoses in low grade glioma patients.

ENLIGHT predicts patient response to targeted and immunotherapies using tumor transcriptomes without prior treatment response data, showing better accuracy than existing transcriptomics-based signatures.

The paper identifies 2,305 SNPs that disrupt 3’ UTR polyadenylation signals and links them to gene expression, eQTLs, and GWAS traits.