Sign in

Hannah Verdonk

@hverdonk.bsky.social
19 followers 9 following 10 posts

Passionate about biology, computers, and the intersection of the two.

PostsRepliesMedia
Reposted by Hannah Verdonk
Genome Biology and Evolution @genomebiolevol.bsky.social · 03/08/2026
@hverdonk.bsky.social, Callan & @sergeilkp.bsky.social present CAPHEINE, a computational workflow that starts with a set of unaligned pathogen sequences and a reference genome and performs a comprehensive exploratory evolutionary analysis. 🔗 doi.org/10.1093/gbe/evag173 #societyjournal #compbio
doi.org
CAPHEINE, or Everything and the Kitchen Sink: A Workflow for Automating Selection Analyses Using HyPhy
Abstract. Here, we present CAPHEINE, a computational workflow that starts with a set of unaligned pathogen sequences and a reference genome and performs a
031
Reposted by Hannah Verdonk
Genome Biology and Evolution @genomebiolevol.bsky.social · 04/08/2026
@sergeilkp.bsky.social @hverdonk.bsky.social et al. present B-STILL, a hierarchical Bayesian framework designed to resolve the selective landscape of protein-coding genes near the zero-rate limit. 🔗 doi.org/10.1093/gbe/evag184 #societyjournal #compbio
doi.org
Beyond Invariable Sites: Using Evolutionary Stasis to Map Multilayered Constraints on the Evolution of Viral and Mammalian Genomes
Abstract. The quantification of genomic conservation has progressed from foundational statistical modeling of evolutionary rates to state-of-the-art deep l
023
Hannah Verdonk @hverdonk.bsky.social · 03/03/2026
The result is a lightweight, functional classifier built in an afternoon — without sacrificing rigor. 🔗 Code: [lnkd.in/e-Wsz5-u](lnkd.in/e-Wsz5-u) AI isn’t replacing computational biologists. Used correctly, it’s a force multiplier.
lnkd.in
LinkedIn
This link will take you to a page that’s not on LinkedIn
010
Hannah Verdonk @hverdonk.bsky.social · 03/03/2026
3️⃣ Review like a scientist, not a spectator. The pipeline worked end-to-end, but small oversights remained (incomplete README examples, rare label handling, re-running predictions). Treat AI as a fast collaborator — not an oracle.
100
Hannah Verdonk @hverdonk.bsky.social · 03/03/2026
2️⃣ Anticipate domain-specific failure modes. Single-cell data has predictable pitfalls: train/test leakage, rare cell types vanishing during splits, and overconfident predictions on unseen states. AI accelerates implementation, but domain expertise prevents subtle errors.
100
Hannah Verdonk @hverdonk.bsky.social · 03/03/2026
1️⃣ Start with structure, not code. I drafted a detailed implementation plan first — specifying donor-aware splits, batch handling, feature alignment, and calibration. AI performs best when operating inside explicit biological constraints.
130
Hannah Verdonk @hverdonk.bsky.social · 03/03/2026
I built a neural cell type classifier in under an hour using AI. The goal wasn’t speed alone — it was to test how models implement workflows from a detailed prompt, the assumptions they make, and what gets lost. Here’s what effective AI-assisted development actually looks like:
110
Hannah Verdonk @hverdonk.bsky.social · 06/08/2025
Just now catching my breath after a fantastic #SMBE2025 a few weeks back. I’m deeply grateful to the local organizers and to @official-smbe.bsky.social for the Graduate Student Excellence Award and for the chance to share my work. It was an honor to present among such brilliant researchers.
081
Hannah Verdonk @hverdonk.bsky.social · 26/06/2025
Thank you very much! It was a great conference.
000
Hannah Verdonk @hverdonk.bsky.social · 24/06/2025
So many great talks from all the students in @lablabella.bsky.social! Let's keep in touch!
000
Hannah Verdonk @hverdonk.bsky.social · 24/06/2025
Really enjoyed your talk @averydb.bsky.social! I'd never considered that frequent transcription of tRNA genes would drive up the local mutation rate, but it makes sense in hindsight.
020
Reposted by Hannah Verdonk
Molecular Biology and Evolution @molbioevol.bsky.social · 24/04/2025
@hverdonk.bsky.social et al. expand on multiclass synonymous substitution (MSS) models, revealing widespread selection on synonymous codon usage across ~12K Drosophila genes. 🔗 doi.org/10.1093/molb... #evobio #molbio #selection
doi.org
A New Comparative Framework for Estimating Selection on Synonymous Substitutions
Abstract. Selection on synonymous codon usage is a well-known and widespread phenomenon, yet existing models often do not account for it or its effect on s
072
Hannah Verdonk @hverdonk.bsky.social · 25/04/2025
I'm honored to receive the Graduate Student Excellence Award and excited to present my research in Beijing this summer!
020