Research

Research

I find a measurement problem, build the tool, and move on.

The common thread across all my work: complex systems that produce signals we don’t yet know how to interpret. A feeding fly. A manuscript under review. A genome-wide RNA modification map. In each case, the scientific question requires inventing the measurement framework first.

Nature Methods β€” in revision

DABEST 2.0

Bootstrap-coupled effect size estimation for multi-group comparisons. Replaces p-values with estimation graphics: mean differences, confidence intervals, and the full sampling distribution of your effect. The delta-delta plot β€” my original contribution β€” handles 2Γ—2 factorial designs and beyond.

β†’ Interactive demo
Science Advances β€” in revision

Ethomics / DESTRA

High-dimensional behavioural state benchmarking in Drosophila. The Espresso platform captures 17 behavioural metrics simultaneously. DESTRA uses bootstrap effect size profiles to ask: does this manipulation produce genuine hunger, or just a superficially similar signature?

β†’ Interactive demo
Current Biology 2017 β€” 48 citations

Intertemporal Choice

Dorsal raphe serotonergic neurons bidirectionally control delay discounting in Drosophila. Activation increases patience; silencing increases impulsivity. Effect is trade-off specific β€” not a motor or motivational confound.

β†’ Project page
In development

LLM Evaluation Pipeline

Adversarial multi-agent pipeline for scientific manuscript evaluation. Four LLM subagents β€” biologist, bioinformatician, reviewer, editor β€” argue toward a structured critique. Building toward bootstrap-estimated inter-rater calibration against human expert reviewers.

β†’ Notebook
PNAS 2020 β€” 119 citations

Noradrenergic Attention

Two distinct noradrenergic pathways implement different attentional control mechanisms in rodents. Selective lesions of each reveal a functional dissociation not apparent from anatomy alone.

β†’ Full publication
Coming soon

Epitranscriptomics

Direct RNA nanopore sequencing pipeline for m6A and pseudouridine mapping. Running on local GPU. Connects back to the behavioural work: does hunger change the RNA modification landscape in serotonergic neurons?

β€’ β€’ β€’

All computational tools are built in Python and released as open-source packages.
DABEST: ACCLAB/DABEST-python