ML researcher at Genentech: generative modeling on manifolds and ML-driven experimental design in biology
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
14
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-16) and cover the author's entire public record, not just the papers listed here. Reproducibility signal comes from Evolve's own reviews.
Research profile
Output—
Not yet scored: this snapshot has no computed output measure. The axis would capture publication volume and consistency; the record does list per-year counts (summing to 80 works over 2017-2026, with 42 in the last three years), but the scored dimension is absent.
Impact—
No impact score is available in this snapshot. The axis would measure citation influence such as citation counts and h-index trajectory; while h-index 14 and i10-index 15 are listed as raw figures, the platform has not scored this dimension.
Focus0/100
The focus axis scores 0, reflecting an even split: the field distribution is 0.5 machine learning and 0.5 computational biology (ML for experimental design). Topic counts spread across five areas — computational drug discovery methods (21), single-cell and spatial transcriptomics (18), GANs and image synthesis (14), adversarial robustness (11), and ML in materials science (11) — indicating methodological breadth rather than a single narrow specialty.
Collaboration51/100
Collaboration scores 51, built on 12 distinct frequent coauthors, including Edward De Brouwer, Carl Edwards, Xiner Li, Namkyeong Lee, and Sara Mostafavi. On the two Evolve-reviewed 2026 papers Scalia appears as a middle author and a last author (neither corresponding), a pattern consistent with team-based work at the machine learning-biology interface rather than solo output.
Activity100/100
Activity scores a maximum 100: per-year counts rise from 5 (2017) through 8 (2021) to 11 (2024), 14 (2025), and 17 (2026), with 42 works across the most recent three years. This is an accelerating, sustained publication cadence that continues into 2026, the strongest signal in the record.
Reproducibility55/100
Reproducibility scores 55, an average computed over the 2 Evolve-reviewed papers, both rated medium readiness. The CRISPR-screen work ships a 1,389-screen benchmark supporting reuse, while the Wasserstein flow-matching paper's entropic optimal-transport map contribution weakens in high dimensions, tempering replication confidence.
Research summary
Gabriele Scalia is a machine learning researcher affiliated with Genentech whose work spans generative modeling theory and machine-learning-driven experimental design in biology. The bibliometric snapshot (OpenAlex identity) records an h-index of 14 and an i10-index of 15, with per-year output rising from 5 works in 2017 to 17 in 2026 and 42 works across the most recent three years — the basis for a maximum activity score of 100. Research interests split evenly (0.5/0.5) between machine learning and computational biology (machine learning for experimental design), which yields a focus score of 0 reflecting breadth rather than narrow specialization; leading topics include computational drug discovery methods (21 papers), single-cell and spatial transcriptomics (18), generative adversarial networks and image synthesis (14), adversarial robustness (11), and machine learning in materials science (11). Two 2026 arXiv papers are reviewed in the Evolve catalog: a middle-author contribution on flow matching in Wasserstein space of Riemannian manifolds, and a last-author study introducing a 1,389-screen CRISPR benchmark with AssayLoop, which attains a 5.67x enrichment factor over random for adaptive hit discovery. Reproducibility averages 55 across two reviewed papers, both rated medium readiness, while collaboration scores 51 on the strength of 12 distinct frequent coauthors.
Collaboration pattern
The record lists 12 distinct frequent coauthors — Edward De Brouwer, Carl Edwards, Xiner Li, Namkyeong Lee, Ehsan Hajiramezanali, Anne Biton, Sara Mostafavi, Doron Haviv, Rishabh Anand, Rex Ying, Aïcha Bentaieb, and Hector Corrada Bravo — matching the collaboration score of 51. Authorship positions on the two reviewed 2026 papers (middle on the manifolds paper, last on the CRISPR-screen paper, neither corresponding) point to team-based projects at the machine learning-biology interface.
Research keywords
Highlights
AssayLoop: amortized adaptive hit discovery across 1,389 CRISPR screens
Flow matching on Wasserstein space of Riemannian manifolds
Accelerating output: 42 works in the last three years
Per-year counts climb from 5 (2017) to 11 (2024), 14 (2025), and 17 (2026), totaling 42 works across 2024-2026 and driving a maximum activity score of 100 in the snapshot.
Timeline
- 2017Earliest year in the per-year record: 5 works, the starting point of a series that climbs steadily through the 2020s.
- 2021Output reaches 8 works, a mid-series high before counts settle at 6 (2022) and 7 (2023).
- 2024Acceleration begins: 11 works in 2024, opening the most productive stretch of the record alongside 14 (2025) and 17 (2026).
- 2026Most productive year with 17 works, including two Evolve-reviewed arXiv papers: Wasserstein flow matching on manifolds (middle author) and the Biology-in-the-loop CRISPR-screen study (last author).
Papers in the Evolve catalog
Frequent co-authors
About this profile
This profile is AI-structured by Evolve.Pub from public bibliographic data (OpenAlex) and Evolve's own paper reviews. It is not written or endorsed by the researcher unless the profile is claimed. Radar dimensions are computed from real bibliometrics with fixed formulas — raw figures are shown alongside; missing dimensions mean insufficient data, never zero. Errors? Claim the profile to correct or hide it.
- Identity and all bibliometrics are name-resolved from a public OpenAlex snapshot; no other sources were used.
- Output and impact axes are unscored in this snapshot; only raw figures such as h-index 14 and i10-index 15 are available.
- The reproducibility mean (55) is computed over just 2 reviewed papers, so it is a small-sample estimate.
- Affiliation (Genentech) and country code (KR) are quoted as given in the source record and have not been independently verified.