Publications

• Pre-trained molecular representations enable antimicrobial discovery

  Authors: Roberto Olayo-Alarcon, Martin K. Amstalden, Annamaria Zannoni, Medina Bajramovic, Cynthia M. Sharma, Ana Rita Brochado, Mina Rezaei, Christian L. Müller

  Antimicrobial resistance is a global threat that necessitates faster discovery of new antibiotics. We introduce a computational strategy using MolE, a deep learning framework, to predict the antimicrobial potential of compounds. Our model identified novel growth-inhibitory compounds, including three new inhibitors of Staphylococcus aureus, offering a promising, cost-effective approach to accelerating antibiotic discovery.

• Development and Validation of Prediction Models for Stroke and Myocardial Infarction in Type 2 Diabetes Patients Based on German Health Insurance Claims Data – Do Modern Machine Learning Methods Outperform Traditional Regression Approaches?

  Authors: Anna Janina Stephan, Michael Hanselmann, Medina Bajramovic, Simon Schosser, Michael Laxy

  Digitalization and big health system data offer new opportunities for targeted prevention and treatment strategies. We developed and validated prediction algorithms for stroke and myocardial infarction (MI) in type 2 diabetes (T2D) patients using German health insurance claims data. Our findings show that claims-based algorithms, using both traditional regression and machine/deep learning methods, perform comparably to existing epidemiological models, with regression-based approaches offering a transparent, scalable, and low-cost solution for cardiovascular risk stratification.

• Learning by doing? The relationship between effort, learning effect and product quality during hackathons of novice teams

  Authors: Nuno Miguel Martins Pacheco, Mara Geisler, Medina Bajramovic, Gabrielle Fu, Anand Vazhapilli Sureshbabu, Markus Mörtl, Markus Zimmermann

  Design education prepares novice designers to solve complex problems with diverse skill sets and an interdisciplinary approach. Hackathons offer hands-on, collaborative learning in a short time frame, emphasizing practical experience and problem-solving. This paper presents an empirical study from a 2-week academic hackathon, examining the relationship between product quality, learning effect, and effort. Data on effort, self-assessed learning, and product quality were collected from 30 teams. The study found a moderate correlation between effort and product quality, but no significant correlation between product quality and learning effect.