FigureAsia 35 Under 35 · Science
Adil Kabylda
Age 29 · Machine-learned molecular simulation · Kazakhstan / Luxembourg
First and corresponding author of SO3LR, a quantum-informed force field trained on four million molecular complexes.
- Approximate age at the edition eligibility date
- 29
- Field
- Computational chemistry
- Country or region
- Kazakhstan / Luxembourg
- FigureAsia U35 Assessment
- 78.7 / 100
Profile
Career and documented record
Molecular simulation usually trades quantum-level accuracy against the ability to model realistic biological scale. Adil Kabylda's 2025 JACS paper introduced SO3LR, a machine-learned force field trained on four million neutral and charged molecular complexes.
The reported system scaled to approximately 200,000 atoms on one graphics processor and was applied to proteins, glycoproteins and lipid bilayers in explicit solvent. On a polyalanine benchmark, the paper reported an eightfold improvement in force error over AmberFF while retaining useful scale.
Kabylda is first and corresponding author and has released supporting data, including work around the QCell resource. SO3LR is enabling research, not a medicine; its importance is that larger biological simulations can be tested against a more expressive learned potential.
FigureAsia selection
Why Adil Kabylda is on the list
Kabylda is selected for a computational platform with unusually clear authorship, scale and open evidence. The ranking credits infrastructure capable of changing biomolecular research—not unproven downstream drug discovery.
Verified work
The 2025–26 record
SO3LR
First- and corresponding-authored a quantum-informed machine-learning force field.
Large-system demonstration
Simulated approximately 200,000 atoms on a single GPU.
Open molecular data
Released data and supporting resources for independent testing and extension.
Field context
The work in its field
Biomolecular simulation is valuable only when a force field remains accurate outside small training molecules and computationally tractable in realistic solvent and membrane environments.
FigureAsia U35 Assessment
Assessment breakdown
78.7out of 100
Substantive 2025–2026 contribution
14.3 / 20
First- and corresponding-authored a quantum-informed machine-learning force field.
Verified scientific impact
11.6 / 15
JACS publication, large-system tests and open data establish a substantial enabling contribution.
Originality and distinction
7.9 / 10
The distinction lies in an equivariant learned potential joining quantum-informed interactions with biomolecular simulation scale.
Field influence
8 / 10
Researchers in machine-learned molecular simulation now have a stronger result to test, extend or challenge because of this contribution.
Individual agency
8 / 10
Kabylda is first and corresponding author, giving him direct responsibility for the central scientific platform.
Durability and trajectory
4.2 / 5
The contribution builds on an active line of work at University of Luxembourg, with further tests and applications still to come.
Asian significance and global relevance
4.2 / 5
Born in Pavlodar, Kazakhstan, and now conducting theoretical chemical physics research in Luxembourg.
Evidential validity and reproducibility
6.4 / 8
Benchmarks and applications are reported openly; accuracy outside represented chemistry remains an explicit limit.
Advance in scientific knowledge
5.6 / 7
The work tests how learned intermolecular physics can transfer from molecular complexes to large solvated systems.
Translational or methodological utility
4.2 / 5
The platform can expand the size and chemical richness of simulations used in molecular science.
Responsible research stewardship
4.3 / 5
Open data improve reproducibility and the profile excludes unsupported claims about drugs or patients.