publications by categories in reversed chronological order. generated by jekyll-scholar.
Research summary
My recent work focuses on practical machine learning for computational biology and healthcare:
Antimicrobial peptide design with diffusion models: We introduced AMP-DiT, a denoising diffusion transformer that generates antimicrobial peptides directly in sequence space, guided by an AMP classifier for activity and diversity. Accepted as a poster at the GenBio Workshop at ICML 2026.
Antibody and protein language models: We introduced ParaAntiProt for paratope prediction, combining antibody-specific signals with protein language model representations to improve antibody binding-site prediction quality.
Federated learning for clinical data: We studied privacy-preserving imputation of missing Parkinson’s disease clinical assessments, showing how federated learning can support robust prediction without centralizing sensitive patient data.
For full citation details, links, and BibTeX entries, see the list below.
2026
GenBio@ICML
AMP-DiT: Antimicrobial Peptide Design with Denoising Diffusion Transformers and Classifier Guidance
AMP-DiT generates antimicrobial peptides directly in sequence space, without relying on protein language models trained on full-length proteins. By guiding generation with an AMP classifier, it produces peptides with strong predicted antimicrobial activity while keeping the generated sequences diverse.
@inproceedings{noroozi2026ampdit,title={AMP-DiT: Antimicrobial Peptide Design with Denoising Diffusion Transformers and Classifier Guidance},author={Noroozi, Alireza and G{\"u}rsoy, Attila},booktitle={GenBio Workshop at ICML},year={2026},url={https://openreview.net/forum?id=KYzd6lcfZ0},}
2025
IEEE JBHI
Bridging the Gaps: Imputation of Parkinson’s Disease Clinical Assessments With Federated Learning
J. Reyes, A. Noroozi, Y. Xiao, and 1 more author
IEEE Journal of Biomedical and Health Informatics, 2025
This paper presents a federated learning approach for imputing missing clinical assessments in Parkinson’s disease datasets, addressing privacy concerns while maintaining prediction accuracy across distributed healthcare institutions.
@article{reyes2025bridging,title={Bridging the Gaps: Imputation of Parkinson's Disease Clinical Assessments With Federated Learning},author={Reyes, J. and Noroozi, A. and Xiao, Y. and Kersten-Oertel, M.},journal={IEEE Journal of Biomedical and Health Informatics},year={2025},publisher={IEEE},doi={10.1109/JBHI.2025.3593459},}
2024
Sci Rep
ParaAntiProt provides paratope prediction using antibody and protein language models
M. Kalemati, A. Noroozi, A. Shahbakhsh, and 1 more author
ParaAntiProt provides paratope prediction using antibody and protein language models. This work presents a novel approach to predicting antibody paratopes using advanced protein language models, contributing to the field of computational immunology and antibody design.
@article{kalemati2024paraantiprot,title={ParaAntiProt provides paratope prediction using antibody and protein language models},author={Kalemati, M. and Noroozi, A. and Shahbakhsh, A. and others},journal={Scientific Reports},volume={14},pages={29141},year={2024},publisher={Nature Publishing Group},doi={10.1038/s41598-024-80940-y},url={https://doi.org/10.1038/s41598-024-80940-y},}