Research

Model Performance Series

June 2026

Astra AI on Enzymes

Enzymes are the catalysts of biology and the most heavily exploited drug-target class in the history of medicine. They are defined not by a shared fold but by the chemistry they perform — which makes them a distinctive computational target.

Çağlar Bozkurt, Aniruddh Goteti · Orbion GmbH · Benchmarked on 6,304 Enzymes

95.3%
Recognized as Enzymatic
ρ 0.93
ΔTm on the T4 Lysozyme Scan
62%
Ligand-Identity Recall

What We Found

Three Results That Matter

Each section of the whitepaper reports the headline performance for one prediction area, its known weaknesses, and where we recommend it for production use.

01

95.3% of the Cohort Recognized as Enzymatic

Classified via the EC head across all seven top-level Enzyme Commission classes, in proportions that track the enzyme universe. Molecular-function GO accuracy reaches 94.2% top-5.

02

F1 Up to 0.89 on PTM Sites

Strongest on myristoylation (0.89), N-linked glycosylation (0.87), and disulfide bonds (0.86). All 39 modification classes are covered, with two operating points per class.

03

ρ 0.93 ΔTm on the T4 Lysozyme Deep Scan

The flagship result of the suite (n = 315). Across the full enzyme set, ρ 0.88 with 90% directional accuracy on strong-effect mutations.

Every number above is reproduced in the whitepaper against public references — Swiss-Prot annotation, PDB co-crystal contacts, and curated experimental thermal-shift data. We report where the models are weak as plainly as where they are strong.

Read the Whitepaper