Gutenberg Open Science
The Open Science Repository of Johannes Gutenberg University Mainz.
Visible. Citable. Open.
Recent Submissions
Item type: Item , Dissertation Access status: Open Access , Improving small molecules activity modelling capability of cell painting data through data augmentation and effective representation learning(2024) Ha, Son V.; Czodrowski, PaulThis thesis focuses on improving image-based activity modeling, for early-stage drug discov ery through data augmentation and representation learning of Cell Painting data. Firstly, a significant contribution is the introduction of the FSL-CP dataset, designed to support few-shot learning (FSL) benchmarking of small-molecule bioactivity prediction using cell microscopy images. Through this dataset we compared several FSL paradigms in a low-data context and study the effectiveness of transfer learning. Additionally, this work proposes an application for underused ‘low concentration images’ in activity modeling. We propose the combination of well-performing models trained at higher image concentrations, with lower image concentration for inference to identify more potent compounds. We show that this approach improves on the conventional method (directly training a high-potency model) in 65% of assays investigated in terms of AUC-ROC, and 75% of assays in terms of RIPtoP-corrected AUC-PR. The thesis further investigates cross-modality representation learning of cell painting (CP) and transcriptomics (TX), which are powerful tools in early drug discovery to gain understanding of the biological effect of compounds on a population of cells post-treatment. In this work, we benchmark two representation learning methods: contrastive learning and bimodal autoencoder. We use the setting of cross modality learning where representation learning is performed with two modalities (CP and TX), but only cell painting is available for new compounds embedding generation and downstream task. This is because for new compounds, we would only have CP data and not TX, due to high data generation cost of the RNA-Seq screen. We show that learned representation improves cluster quality for clustering of CP replicates and different modes of action (MoA). clustering of CP replicates and different modes of action (MoA).Item type: Item , Dissertation Access status: Open Access , Item type: Item , Zeitschriftenaufsatz Access status: Open Access , Ultrafast dynamics of chiral spin structures in synthetic antiferromagnets(2025) Guo, Zongxia; Gruber, Raphael; Ksenzov, Dmitriy; Léveillé, Cyril; Pancaldi, Matteo; Pedersoli, Emanuele; Spezzani, Carlo; De Ninno, Giovanni; Capotondi, Flavio; Gutt, Christian; Kläui, Mathias; Cros, Vincent; Reyren, Nicolas; Jaouen, NicolasIn synthetic antiferromagnetic multilayers (SAFs), chiral magnetic structures such as spin spirals and skyrmions have been stabilized at room temperature by precisely tuning the effective perpendicular magnetic anisotropy, the Dzyaloshinskii-Moriya interaction, and the Ruderman-Kittel-Kasuya-Yoshida (RKKY) interlayer coupling. In this study, we investigate the dynamics of spin spirals on ultrashort timescales after femtosecond laser pumping in SAFs. The access to ultrafast magnetization dynamics, inaccessible by conventional optical techniques due to the zero net magnetization, has been enabled by the use of time-resolved circular dichroism in x-ray resonant magnetic scattering (CD-XRMS). A pair of two-dimensional x-ray scattering patterns for left and right elliptical polarization (EL and ER) have been recorded for each delay. In contrast to our previous findings in ferromagnetic multilayers, the magnetization (EL+ER) and dichroism (EL−ER) signals exhibit notably similar ultrafast dynamics, with demagnetization occurring on a timescale of ∼180 fs, followed by rapid remagnetization within ∼500 fs. This similarity in ultrafast dynamics can be attributed to the continuous rotation of magnetization in the spin spiral of SAFs, which evolves smoothly in space without forming abrupt domains or alternating domain walls. The ultrafast response and stability in its topological character highlight the potential of SAF-based chiral magnetic structures for future high-speed, energy-efficient data storage and processing applications.