DELPHAI predicts heterogeneous perturbation responses with learned single-cell fitness

Preprint (not peer reviewed; no independent replication found) from aiPTO TechBio authors: DELPHAI trains a fitness network and an optimal-transport network jointly and, at inference, applies fitness-gated transport with direct gene-space retrieval. The abstract self-reports ranking first on differentially expressed gene prediction across two benchmark frameworks. Authors are affiliated with aiPTO. Same-name successor of VC0129 (2025 preprint) but a different method.

Authors / organisation

Zhang X, Wu H, Liu H (aiPTO TechBio GmbH; Liu also DKFZ)

Venue

bioRxiv (preprint; v1 2026-07-06, v2 2026-09-17)

Date

2026-07-06

Type

preprint

Year

2026

DOI

10.64898/2026.07.01.735965

Verification

Crossref API + bioRxiv API (title, authors, affiliations, v1 date 2026-07-06, v2 date 2026-09-17, DOI, abstract) checked 2026-10-17; full text and competing-interest statement not read

PerturbationBenchmark & validationDrug responseDigital twin

Open in the searchable library → · JSON