← Back to home

Computational biology and machine learning for drug discovery, on Kallyope's gut-brain biology programs. I worked there from Series A through Series C, as the programs moved from preclinical discovery into the company's first human trials.

Single-cell atlas of a novel tissue

I led the computational analysis for a single-cell atlas of a tissue that had not been characterized at this resolution, working with wet-lab scientists to define cell types via graph-based clustering. Choosing the clustering resolution was part of the work: the useful answer is the one a pharmacologist can act on, which is a decision made with the biologists rather than handed to them. The results contributed to a $2M milestone with Novo Nordisk.

A public fetal human gut atlas (Elmentaite et al. 2020, Developmental Cell, via CZ CELLxGENE Discover, CC-BY), every cell placed in three dimensions by transcriptional similarity and shaded by body system. Not Kallyope data: nothing proprietary appears here.

From preclinical discovery to first human trials

I developed the analysis and machine-learning pipelines used across several discovery programs, contributing to identification of the lead compounds that reached the company's first human clinical trials.

I also presented research progress to the Scientific Advisory Board, which included three Nobel Laureates.

Mapping neural circuits

I co-developed probabilistic models that map neuronal connections by combining viral tracing with single-cell sequencing. Tracing indicates which regions are connected; sequencing identifies cell types. Reconciling the two is a modelling problem, and the models were built to infer circuit architecture from both measurements together.