The Lavinda Lab · Chemistry-Aware AI Validation

When can you trust an AI-predicted structure?

AI can now generate protein-complex hypotheses in seconds. The hard part is no longer making the prediction. It is knowing which predictions are chemically real. My lab builds chemistry-aware methods to tell them apart, in the cases where metals, glycans, cofactors, and interface energetics decide the answer.

Olga Lavinda, PhD · Principal Investigator
Clinical Assistant Professor of Chemistry · Stern College for Women, Yeshiva University
Katz School of Science and Health · NIH NRSA-trained (NYU Langone)

← lavinda.ai ORCID Scholar YU Faculty

"AI predictions are cheap now. Trustworthy biochemical interpretation is not. That gap, not more prediction, is where my lab works."

Current projects

One question, several systems.

One question runs through all of it: when does an AI-predicted structure reflect real biochemistry, and when is it just a confident guess? Every project tests that where chemistry matters most, and pairs computation with experiment.

Structural biology · AI validation

Can you trust an AI-predicted protein complex?

AlphaFold 3 proposes a protein assembly in seconds, but not every prediction is real. We test it where it tends to fail: metal ions, sugar chains, and the randomness from one run to the next. Our system is the melanogenic enzyme complex (TYR-TYRP1-TYRP2), whose assembly underlies albinism and is a drug target in melanoma. Across hundreds of runs the result is consistent: a high confidence score and a structure that reproduces are not the same thing. Telling them apart is the work.

The predicted TYR-TYRP1-TYRP2 melanogenic trimer (AlphaFold 3).

A confident score is not a reproducible structure, across hundreds of AF3 runs — Each point is one model run. A high confidence score (right) does not guarantee a structure that reproduces (top). That gap is what we measure.

Interface reproducibility across full-state seeds, by interface — Read straight off the data: one interface reproduces reliably (TYRP1-TYRP2), one is a coin toss (TYR-TYRP2), one never holds (TYR-TYRP1). A biological signal, not a modeling artifact.

Enzyme mechanism · Histamine biology

What in your diet blocks histamine breakdown?

Histamine intolerance affects a large, under-served population, yet the enzyme that clears dietary histamine, human diamine oxidase (hDAO / AOC1), is barely characterized as a diet or drug target. We use AI structure prediction and docking to rank which everyday dietary compounds interfere with it, including terpenes from common herbs and oils that no one has tested against hDAO. This is the focus of our current NIH R15 application. AlphaFold gives us a reliable structure to dock into, and compound screening and enzyme assays are running now.

AF3 folds human DAO with high, reproducible confidence across all five models — A consistent, high-confidence AF3 fold of hDAO across all five models. A reliable structure to dock into.

Protein design · ML for structure

A grammar of protein stabilization.

Some protein surfaces are strained and unstable; others lock cleanly into complexes. We are looking for the reusable rules that separate the two: a grammar of stabilization drawn from PDB structures, binding energetics, and protein language models. If those rules transfer, they tell us which proteins will pair stably, and why.

Energetic Dialects: frustrated surfaces to stabilization grammar to stabilized complex

Biocatalysis · Green chemistry

Cleaner ways to oxidize stubborn hydrocarbons.

How do enzymes selectively oxidize some of the most inert hydrocarbons, and can they do it in greener solvents? We study fungal peroxygenases and their active-site chemistry for selective C-H oxyfunctionalization of branched and cyclic alkanes in deep eutectic solvents, with an eye toward cleaner chemical manufacturing.

Fungal peroxygenase selectively oxidizes an inert C-H bond of a cyclic alkane using hydrogen peroxide in a deep eutectic solvent — One inert C-H bond, selectively oxidized under mild, greener conditions: hydrogen peroxide as the oxidant, a deep eutectic solvent as the medium.

Methods

An open, chemistry-aware validation pipeline.

One pipeline runs under every project: an open AlphaFold 3 validation workflow built for the cases where chemistry decides the answer, metals, glycans, cofactors, multi-seed reproducibility, and interface energetics. It is how the lab separates a confident prediction from a chemically real one, and it pairs computation with wet-lab measurement at every step.

Computational

AlphaFold 3 multi-seed prediction

Ensemble runs with curation and reproducibility scoring, never a single-shot prediction.

ML-augmented molecular dynamics

Stability and conformational sampling beyond the static fold.

QM/MM active-site mechanism

Quantum treatment of the reactive center in its protein environment.

Protein-ligand docking

AutoDock Vina for compound screening and allosteric-site exploration.

Interpretable ML & protein language models

Structure-function rules that transfer, drawn from PDB structures and sequence embeddings.

Wet-lab & biophysical validation

Enzyme kinetics

Functional assays that test what the models predict.

DSC / thermal stability

Differential scanning calorimetry for fold stability.

ITC binding

Isothermal titration calorimetry for interaction energetics.

Reproducible pipelines

Pre-registered predictions and automated metadata throughout.

Presentations

Four talks at ACS Fall 2026.

Presenting across four ACS divisions in one meeting, spanning structural biology, medicinal chemistry, computational science, and chemical education, including an invited symposium and a Sci-Mix selection.

Deep Learning Co-Folding Models versus Computational Physics-Based Models of Protein-Ligand Interactions

COMSCI Innovative Program · invited · Committee on Science symposium. Oral session and integrated digital-screen poster.

Teaching with AI Without Losing the Student

CHED · selected for Sci-Mix · a structured framework for AI-literacy integration in undergraduate chemistry.

The Prompt Ladder AI-literacy framework graphical abstract

AlphaFold 3 Reveals the Architecture of the Melanogenic Enzyme Complex

BIOL · Biological Chemistry poster · metal coordination, glycosylation, and the structural basis of oculocutaneous albinism. Lavinda, Aharon, Cohen, Dube & Muyambo.

Melanogenic complex TYR-TYRP1-TYRP2 graphical abstract

Free Docking and Allosteric Site Exploration of Breastfeeding-Supplement Ingredients at Human Diamine Oxidase

MEDI · Medicinal Chemistry poster · an AlphaFold 3 and AutoDock Vina study.

hDAO breastfeeding-supplement docking graphical abstract

Papers

Peer-reviewed.

Origin of high diastereoselectivity in reactions of seven-membered-ring enolates

Lavinda, Witt & Woerpel · Angewandte Chemie · 2022

Biophysical compatibility of the tyrosinase-TYRP1-TYRP2 metalloenzyme complex

Frontiers in Pharmacology · 2021

GCN2 kinase inhibitors

Computational and Structural Biotechnology Journal · 2018

Thermochromism of indigo and Tyrian purple

Dyes & Pigments

Full publication list → In preparation

Chemistry-aware benchmarking of AF3 for metalated, glycosylated complexes

When AlphaFold confidence metrics succeed or fail: multi-seed reproducibility, metal and glycan handling, and interface-energetics descriptors.

Structural model of the TYR-TYRP1-TYRP2 complex and OCA genotype-phenotype

In-silico mutagenesis of albinism-associated variants at the complex interface versus the active site.

Training

A program that turns students into scientists.

The lab runs a structured, seven-phase program that takes students with no computational-biology background and moves them, one skill-gated phase at a time, to real work on unpublished problems, conference posters, and co-authorship. It is a system, not ad-hoc apprenticeship: a written manual, a personal electronic notebook for every student, standardized data and quality-control practice, and a mentorship loop with sign-off at every phase.

Background & framing

Structure & sequence acquisition

Visualization & validation

Quality assessment (MolProbity)

AI structure prediction (AlphaFold 3)

Interface & reproducibility analysis

Figures & communication

Advanced tracks add AutoDock Vina docking and wet-lab enzyme kinetics. The principle is depth before automation: students do each step by hand and learn to judge an output before they are trusted with the pipeline that produces it. Rigor is a first-class skill here, with pre-registered predictions, blinded scoring, and language that never overstates a result.

By the end, a student can take a project the whole way: from the first paper, through structure prediction and validation, to a conference poster they present under their own name.

Interested in joining? Undergraduate and graduate researchers can apply below. No prior computational-biology experience required.

Apply to the lab →

Team

The people doing the work.

The lab trains undergraduate women researchers at Stern College for Women, near-peer co-mentored by MS Biotechnology graduate researchers, a structure that widens access to computational science. The undergraduate cohort is growing, and prospective students are welcome to inquire.

New York City and the classroom — The lab runs on New York City, a lot of coffee, and students who show up. Working with students at premier NYC colleges keeps the science honest.

Olga Lavinda, PhD · Principal Investigator

Scientific direction, AI-workflow and validation-framework design, computational structural biology, and student supervision. ACS faculty mentor, Yeshiva University student chapter.

Graduate researchers · MS in Biotechnology, Katz School of Science and Health

Nicolas Hove

hDAO protein-ligand modeling and compound prioritization.

Tinashe Rabson Muyambo

TYR-TYRP1-TYRP2 AlphaFold 3 benchmarking and interface reproducibility.

Ntobeko Dube

TYR structural analysis and molecular visualization.

Meli Nkau

Research data integration and visualization across projects.

Londiwe Moyo

hDAO and HNMT structure prediction.

Undergraduate researchers · Stern College for Women

Tiferet Aharon

TYR complex modeling and OCA variant analysis.

Netanya Cohen

TYR complex modeling and interface analysis.

Dalit Gulkarov

hDAO docking and structure prediction.