From correct analyses to decisions: reading OncologyBench
An audit of two public examples, seven original metadata fixtures and a hypothetical value-of-information calculation distinguish evidence scope from project decisions.
Daiki Kumakura
I am a mathematical biologist and a bioinformatics researcher in the pharmaceutical industry. I received my Ph.D. in Life Science from Hokkaido University in March 2025.
I study biological dynamics using mathematical models and data analysis, with an interest in pharmacometrics and systems biology.
ORCIDGoogle ScholarLinkedInGitHub
Recent writing
Can evaluation metrics distinguish a technical duplicate from a mean baseline?
An independent audit of five metrics on three public Perturb-seq datasets, comparing technical duplicates with a mean baseline without evaluating predictive models.
Can public data show that ivonescimab 20 mg/kg Q3W beats 10 mg/kg Q3W?
A Japanese research note auditing what published phase 1-2 pharmacokinetics, PD-1 receptor occupancy and dose-level response rates can say about 20 versus 10 mg/kg Q3W of ivonescimab, with the analysis plan fixed before computing. The answer is inconclusive. It makes no claim about efficacy ranking, recommended dose or any regulatory decision.