Ahmet Sarigun

About

I am Scientist at the Max Delbrück Center for Molecular Medicine (MDC-BIMSB) at Bioinformatics & Omics Data Science Platform in Berlin, Germany.

I currently work on Structure Based Drug Design, with a focus on:

• Computational Chemistry (Force Field)
• Deep Learning (Geometric Deep Generative Models)
• High Throughput Virtual Screening & ADME

Background

• Education: BSc in Chemistry from METU (Middle East Technical University), Department of Chemistry, Ankara, Turkey, 2023
• Research Experience:
  • Scientist (2024-Present): Structure Based Drug Design at Max Delbrück Center for Molecular Medicine (MDC-BIMSB), Bioinformatics & Omics Data Science Platform, Berlin, Germany
  • Research Project (2023): Graph Representation Learning for Chemistry supervised by Antoine Marion, METU, Ankara, Turkey
  • Research Project (2021-2023): Graph Representation Learning supervised by Ahmet Rifaioğlu (2021-2023) and Dominique Beaini (2021-2022), Remote
  • Summer Intern (2022): Deep Learning for Single Cell and Spatial Transcriptomics, supervised by Ahmet Rifaioğlu and Jovan Tanevski, Heidelberg University Hospital, Institute for Computational Biomedicine, Heidelberg, Germany
  • Research Intern (2021-2022): de novo Drug Design using Adversarial Machine Learning supervised by Tunca Doğan, Hacettepe University Biological Data Science Laboratory, Ankara, Turkey

Other Interests

Educational Animations

I sometimes create educational animations in Quantum Chemistry using manim. Check out my content on YouTube.

Quantum Chemistry | Waves and Particles - A quick video that starts of a series of videos on quantum chemistry with waves and particles that we are going to go about understanding the ideas through fun animations.

Running & Cycling

As an avid runner and cyclist since high school, I enjoy regular & irregular runs and rides.

More running and biking routes can be found on Strava.

Research

Publications

CompassDock: Comprehensive Accurate Assessment Approach for Deep Learning-Based Molecular Docking in Inference and Fine-Tuning

• The CompassDock framework is a comprehensive and accurate assessment approach for deep learning-based molecular docking. It evaluates key factors such as the physico-chemical properties and bioactivity favorability of ligands.
  A Sarıgün, V Franke, B Uyar, A Akalin
  • Oral Presentation at Helmholtz AI Conference, 2024, Düsseldorf, Germany
  • Accepted to NeurIPS MLSB, 2024, Vancouver, Canada
  • [arXiv, 2024] | [GitHub Repo] | [PyPI Installation] | [Colab Demo]
  

Flexynesis: A deep learning framework for bulk multi-omics data integration for precision oncology and beyond

• A deep-learning based multi-omics bulk sequencing data integration suite with a focus on (pre-)clinical endpoint prediction.
  B Uyar, T Savchyn, R Wurmus, A Sarıgün, MM Shaik, V Franke, A Akalin
  • [bioRxiv, 2024] | [GitHub Repo]
  

Target Specific De Novo Design of Drug Candidate Molecules with Graph Transformer-based Generative Adversarial Networks

• A new deep learning model architecture for de novo drug design to treat drug-resistant diseases using Generative Adversarial Networks and a combination of transformer and graph neural network approaches.
  A Ünlü, E Çevrim, A Sarıgün, MG Yiğit, H Çelikbilek, O Bayram, HA Güvenilir, A Koyaş, DC Kahraman, A Olğaç, AS Rifaioğlu, E Banoğlu, T Doğan
  • [arXiv, 2023] | [Main GitHub Repo] | [Old GitHub Repo] | [Hugging Face Demo]
  

Graph Mixer Networks

• A novel framework that incorporates the principles of MLP-Mixers for graph-structured data.
  A Sarigun
  • [arXiv, 2023] | [GitHub Repo] | [Blog/Story]
  

Multi-Mask Aggregators for Graph Neural Networks

• New aggregation method for GNNs, where the model learns a weighted mask for each aggregator before collecting neighboring messages.
  A Sarigun, AS Rifaioglu
  • Accepted to 1^st Learning on Graphs Conference, 2022
  • [GitHub Repo] | [Paper] | [Poster]
  

Multimodal single cell data integration challenge: results and lessons learned

• NeurIPS 2021 Multimodal single cell data integration challenge.
  C Lance, et. al | Consortium Author
  • Accepted to Proceedings of Machine Learning Research, 2022
  • [Blog] | [GitHub Repo]

Presentations

• COMPASS: Enhancing Deep Learning Based Molecular Docking Insights for Comprehensive Analysis, Helmholtz AI Conference 2024 (June 2024) , Düsseldorf, Germany
• FEXBind: Fast Explainable Binding Free Energy Prediction by Leveraging Molecular Mechanic Features and Molecular Structure with Graph Neural Networks, Undergraduate Research Project (June 2023) Ankara, Turkey
• Multi-Mask Aggregators for Graph Neural Networks, 1^st Learning on Graphs Conference (December 2022), Online
• Discovery of Functional Motifs by Association to Clinical Features with GNNExplainers, Heidelberg University Hospital, Institute for Computational Biomedicine (September 2022), Heidelberg, Germany
• De novo Drug Design using Deep Generative Models, Hacettepe University BioData Lab (September 2021), Ankara, Turkey
• TransGAN: Two Transformers Can Make One Strong GAN, CENG 796 Deep Generative Models (May 2021), Online | [YouTube Video] [Slides] [Unofficial Implementation]

Contact

• Twitter
• GitHub
• LinkedIn
• YouTube
• Mastodon

Elements

i = 0;

while (!deck.isInOrder()) {
    print 'Iteration ' + i;
    deck.shuffle();
    i++;
}

print 'It took ' + i + ' iterations to sort the deck.';