PhD Scientific Days 2025
Budapest, 7-9 July 2025
Poster Session II. - J: Theoretical and Translational Medicine
Matrix Transformation and Pre-Processing Techniques to Improve the Performance of a Generative Artificial Intelligence Model Predicting Protein–Protein Interactions from Network Topology
Előadó neve
Mr. Balogh Olivér, MSc
Neptun code
HYKT8D
Előadó munkahelye
Semmelweis University Department of Pharmacology and Pharmacotherapy
Előadó telefonszáma
06203685079
Előadó e-mail címe
balogh.oliver.marton@semmelweis.hu
Az előadás címe
Matrix Transformation and Pre-Processing Techniques to Improve the Performance of a Generative Artificial Intelligence Model Predicting Protein–Protein Interactions from Network Topology
Szerző(k) neve és munkahelye
Balogh Olivér1, Sagát Lóránt1, Horváth András2, Ferdinandy Péter1, Ágg Bence1
1: Cardiometabolic and HUN-REN-SU System Pharmacology Research Group, Department of Pharmacology and Pharmacotherapy, Semmelweis University, Budapest
2: Faculty of Information Technology and Bionics, Pázmány Péter Catholic University, Budapest
Bemutatás módja
Poszter
Szekció
Poster Session II. - J: Theoretical and Translational Medicine
Language of the presentation
English
Preferred session
Theoretical and Translational Medicine
Összefoglaló szövege
Introduction
Previously, we introduced a generative artificial intelligence (AI) model for protein-protein interaction prediction using PPI networks (interactomes). This model handles extracted regions (subgraphs) of the interactome similar to images, in an image-to-image translation framework.
Aims
Our aim is to improve the performance of our AI model by introducing various pre-processing techniques for the augmentation of the input subgraphs.
Methods
Bandwidth optimization algorithms were used to create feature-rich structures in the adjacency matrix representation of the subgraphs. Gaussian noise and mirroring/rotation were used to increase input data variance. Hyperparameters of the model were adjusted using literature search and empirical testing. Model performance was evaluated on interactomes from multiple sources, such as the STRING (Search Tool for the Retrieval of Interacting Genes/Proteins) database, using metrics like area under the precision-recall curve (AUPRC) and precision at 500 (P@500).
Results
By our large-scale evaluation on publicly available PPI networks, the best performing combination of pre-processing techniques resulted in significant improvements. For example, the AUPRC values increased from 0.103, 0.119, and 0.095 to 0.396, 0.444, and 0.377 on the variants of the Homo sapiens interactome filtered by confidence thresholds 0.7, 0.9 and 0.95, respectively.
Conclusion
In conclusion, we demonstrate the performance improving effects of different matrix pre-processing techniques on a generative AI model performing topology-based PPI prediction. Our approach further facilitates the use of network theory and AI for drug target identification.
Funding
Project no. RRF-2.3.1-21-2022-00003 has been implemented with the support provided by the European Union. This project has received funding from the HUN-REN Hungarian Research Network. The research was financed by the Thematic Excellence Programme (2020-4.1.1.-TKP2020) of the Ministry for Innovation and Technology in Hungary, within the framework of the Therapeutic Development and Bioimaging thematic programmes of the Semmelweis University. This study has been supported by a Semmelweis 250+ Excellence Fellowship and by the EKÖP-2024-23 New National Excellence Program of the Ministry for Culture and Innovation from the source of the National Research, Development and Innovation Fund.
University
Semmelweis University
Supervisor
Bence Ágg, MD, PhD
Publication of my abstract
I do not give consent to the publication of my abstract on the website of the congress.
phd.section.field
in doctoral studies after complex exam (PhD)
Kind
Szabad
Status
elfogadva
Accepted presentation method
poszter
Előadás fájl jóváhagyás
nem rendelkezett róla
Előadó
7403
Start
19:06
End
19:12