Saeed Rafieyan

Project · 2022–present · Saeed Rafieyan, lead developer

MLATE

Machine Learning Applications in Tissue Engineering: publicly available models and data that predict how a scaffold will print and how cells will respond to it, rank candidate formulations, and suggest what to fabricate next.

Role
Lead developer
Methods
Gradient boosting, deep and tabular foundation models, Bayesian optimisation, LLM-assisted protocol generation
Data
Literature-derived scaffold dataset, published with the code
The MLATE V3 pipeline. Phase A is built once from the literature corpus, including data gathering, data imputation, clustering, classification, and WSSQ definition. Phase B runs for each user query, including web app development, bioink optimization, and protocol generation.
Fig. 1. From literature to dataset to models (MLATE V3).

Versions

VersionYearScopePaper
V12023Cardiac scaffolds; 28 algorithms benchmarked for cell behaviourComputers in Biology and Medicine
V220243D (bio)printed scaffolds; 1,171 scaffolds, 60 biomaterials, 49 cell linesBiofabrication
V32026Multi-tissue; WSSQ ranking, Bayesian optimisation and a public web appBiofabrication

About

Designing a tissue-engineering scaffold means choosing biomaterials and their concentrations, cell types and densities, and printing conditions, then running experiments to see what works. Most of that search is still trial and error. MLATE turns the published literature into structured datasets and trains models that predict the outcome from variables known before fabrication.

What it does

  • Predicts cell response and printability from pre-fabrication variables. In V3 the best models reach weighted F1 scores of 0.773 (cell response) and 0.806 (printability).
  • Ranks candidates with the Weighted Synergistic Scaffold Quality (WSSQ) score, which combines both predictions with adjustable weights.
  • Optimises formulations and printing parameters inside a user-defined design space with Bayesian optimisation.
  • Plans experiments: an LLM-assisted module drafts step-by-step fabrication protocols for the shortlisted candidates.

The aim is to narrow the experimental search space and suggest candidates worth testing, not to replace experimental validation.

Try it

The web app runs at MLATE, and the code, datasets and analysis pipeline are on GitHub.

How to cite

Please cite MLATE V3: An Open-Source Multi-Tissue AI Framework for Data-Driven Optimization of 3D-Printed and Bioprinted Scaffolds.

@article{rafieyan2026mlate,
  title     = {{MLATE V3: An Open-Source Multi-Tissue AI Framework for Data-Driven Optimization of 3D-Printed and Bioprinted Scaffolds}},
  author    = {Rafieyan, Saeed and Partovi Nasr, Minoo and Ansari, Elham and Kheradvar Kolour, Alireza and Banimohamad-Shotorbani, Behnaz and Vasheghani-Farahani, Ebrahim},
  journal   = {Biofabrication},
  year      = {2026},
  month     = oct,
  doi       = {10.1088/1758-5090/aeb162},
  note      = {In press}
}