01/13/2026
š¬New Publication in Nature Portfolio led by Elliot, Lingxin and Jack Shireman
š§ š§¬ We can read thousands of genes in a single cellā¦
but still struggle to explain what that cell actually is.
That gap is now one of the biggest barriers to turning data into patient care.
So we built CASSIA.
CASSIA is an AI system that works more like a clinical team than a black box.
Multiple AI agents analyze cell data, challenge each other, andāmost importantlyāexplain their decisions.
Why this matters ā¬ļø
In cancer, immunotherapy, and neuro-oncology, identifying therapy-resistant tumor cells or critical immune populations can change how we:
⢠design treatments
⢠predict recurrence
⢠move discoveries toward patients
AI thatās accurate but not interpretable isnāt enough.
Our goal isnāt to replace scientists and clinicians. It is to scale expertise, reduce bias, and make single-cell biology usableānot mysterious.
š¬ If AI is going to belong in medicine, it must be transparent and trustworthy.
We believe CASSIA is a step in that direction. What would make you trust AI in clinical decision-making?
Assigning cell types in single-cell RNA-seq is essential yet challenging, as it requires expertise, time, and is often subjective. Here, the authors present CASSIA, a multi-agent AI system that provides automated, interpretable, and quality-controlled annotations with high accuracy.