Multimodal & Computational Biology
I’m Dean Tessone. I build representation and foundation models that connect cancer biology from DNA and RNA, to single cells, tissue, and the whole patient.
Explore the research ↓
01 / 05 · MOLECULAR
At Tempus AI, I build unimodal representation encoders for cancer genomics and transcriptomics, then integrate those representations with other clinical modalities.
Foundation-model representations for genomic variation and transcriptomic state.
DNA · RNA · CANCER BIOLOGY02 / 05 · ONE CELL
A deep contrastive-learning framework extracts robust features from four-channel single-cell immunofluorescence images for phenotyping across whole-slide liquid-biopsy data.
SCIENTIFIC REPORTS · 2025 Robust single-cell phenotypingContrastive feature learning for classification, clustering, outlier detection, and enumeration of circulating-cell phenotypes.
Read the paper ↗03 / 05 · BLOODSTREAM
The RED algorithm trains a slide-specific denoising autoencoder on approximately 2.5 million IF tiles and uses reconstruction error to rank events by rarity.
NPJ PRECISION ONCOLOGY · 2025 Rare events in liquid biopsyUnsupervised anomaly detection that reduces each slide to a ranked cohort of 2,500 rare tiles for downstream review.
Read the paper ↗04 / 05 · TISSUE
The tissue view pairs H&E morphology with spatial proteomics. My OpenIMC work focuses on unified, reproducible analysis of imaging mass cytometry data.
BMC BIOINFORMATICS · 2026 OpenIMCIntegrated GUI and command-line workflows for IMC visualization, preprocessing, segmentation, phenotyping, and spatial analysis.
Read the published paper ↗05 / 05 · ONE PATIENT
At Tempus AI, I work on a unified system that integrates pathology, genomics, transcriptomics, and ctDNA to model a patient’s tumor biology.
Learning a patient representation across genomics, transcriptomics, circulating tumor DNA, and tissue morphology.
DNA + RNA + ctDNA + PATHOLOGYResearch scope
Models across scales:
molecules to patients.
Trajectory
My Ph.D. work at USC spans contrastive representation learning for single-cell IF, denoising-autoencoder rare-event detection, and reproducible IMC analysis. At Tempus AI, I build unimodal encoders for cancer genomics and transcriptomics and help extend them into multimodal foundation models spanning pathology, DNA, RNA, and ctDNA.
Download curriculum vitae ↗Selected papers
Field notes
Research reflections, side projects, and things I’m learning will live here.
LAB NOTEBOOK · COMING SOONContact