Dr. Anna Kahler

Dr. Anna Kahler

User Support Life Science

Central Scientific Institutions
Erlangen National High Performance Computing Center

Room: Room 1.036
Martensstraße 1
91058 Erlangen

My current focus is on finding optimized runtime parameters for GROMACS on GPU and CPU clusters. So far, I have blogged about Multi-GPU Gromacs Jobs on TinyGPU, the performance of GROMACS on Intel Xeon Ice Lake vs. NVIDIA A100, A40, and GROMACS performance on different GPU types. New posts about GROMACS performance on modern GPU hardware, ARM architectures, and AMD CPUs are still in progress.

I am also helping users with difficult simulation setups and some of these support cases are briefly summarized: Multi GPU setup in GROMACS for large simulation, Non-default REMD setup with GROMACS on multi GPU, and The power of modern GROMACS versions on modern hardware.

 

List of Publications

2016

2015

2013

  • Horn ACH, & Kahler A, Letter to the Editor: The effect of fulvic acid on pre- and postaggregation state of Aβ17-42: molecular dynamics simulation studies, S. Verma, A. Singh and A. Mishra, Biochim Biophys Acta 1834 (2013) 24-33. Biochimica Et Biophysica Acta-Proteins and Proteomics 2013, 1834(12), 2867-8. https://dx.doi.org/10.1016/j.bbapap.2013.07.012
  • Kahler A, Sticht H and Horn AHC, Conformational Stability of Fibrillar Amyloid-Beta Oligomers via Protofilament Pair Formation – A Systematic Computational Study, PLoS ONE 2013, 8, e70521. https://doi.org/10.1371/journal.pone.0070521

Dissertation

Intrinsic Flexibility and Structural Stability of Proteins

An open access version of the thesis is available from OPUS FAU.

Abstract

The properties of a protein are based on its fold and thus, based on the sequence of the amino acids. Molecular and structural biology have provided a wealth of information about proteins, however, the underlying dynamical processes are not yet fully understood. In this context, molecular dynamics (MD) simulations represent a computational method that is capable to supplement experimental data by providing atomistic information on the dynamical behavior of proteins. This thesis illustrates how MD simulations can be applied to investigate biologically relevant proteins on the molecular level.

Transcription factors control the flow of genetic information from DNA to messenger RNA rendering understanding of their function highly important for the modulation and optimization of protein expression. The transcription factor RfaH from Escherichia coli consists of an N-terminal domain (NTD) and a C-terminal domain (CTD), which tightly interact in the autoinhibited conformation of RfaH. Upon activation, the CTD is released and undergoes a large-scale α → β structural transition. In the present work, investigation of RfaH under different environmental conditions revealed that not only high temperatures, but also a decrease in ionic strength significantly enhances CTD dynamics. Despite this enhanced dynamics, none of the conditions investigated caused CTD dissociation suggesting that this process needs to be triggered by the interaction with DNA or other proteins of the transcription machinery. Further, the N-terminus of the first CTD helix, which contains two glycines, was identified to exhibit rather large motions and a set of mutations is proposed affecting the local dynamics and/or the strength of the NTD–CTD interactions.

G-protein coupled receptors (GPCRs) interact with small molecules, peptides, or proteins and transmit a signal over the membrane via structural changes to activate intra-cellular pathways. GPCRs are characterized by a rather low sequence similarity and exhibit structural differences even for functionally closely related GPCRs. Here, a computational approach is proposed that relies on the generation of several independent models based on different template structures, which are subsequently refined by MD simulations. The conformational stability and the agreement with GPCR-typical structural features is then used to select a favorable model. This strategy was applied to predict the structure of the herpesviral chemokine receptor US28 by generating three independent models based on the known structures of the chemokine receptors CXCR1, CXCR4, and CCR5. Model refinement and evaluation suggested that the GPCR-typical structural features, such as a conserved water cluster or conserved non-covalent contacts, are present to a larger extent in the model based on CCR5 compared to the other models. A final model validation based on the recently published US28 crystal structure confirms that the CCR5-based model is the most accurate and exhibits 80.8 % correctly modeled residues within the transmembrane helices. The same modeling strategy was used to investigate the structure for the herpesviral chemokine receptor US27; refinement and evaluation of the models suggested that the model based on CXCR4 exhibits more GPCR-typical structural features than the other models. This favorable model for US27 was afterwards used to elucidate the interface of US27 and the CXCR4. The data suggest that the interaction between the heterodimer is stronger than in the CXCR4–CXCR4 homodimer which is in line with previous research that strong host-pathogen interaction are required to interfere with host signaling processes.