A5027895655- Protein Structure and Dynamics
- Computational Drug Discovery Methods
- Ion channel regulation and function
- Mass Spectrometry Techniques and Applications
- Machine Learning in Materials Science
- Receptor Mechanisms and Signaling
- Cardiac electrophysiology and arrhythmias
- Manufacturing Process and Optimization
- Spectroscopy and Quantum Chemical Studies
- Enzyme Structure and Function
- Neuroscience and Neuropharmacology Research
- RNA and protein synthesis mechanisms
- Robotic Path Planning Algorithms
- Electrochemical Analysis and Applications
- Epilepsy research and treatment
- Lipid Membrane Structure and Behavior
- AI-based Problem Solving and Planning
- thermodynamics and calorimetric analyses
- Inertial Sensor and Navigation
- Neuroscience and Neural Engineering
- Bioinformatics and Genomic Networks
- Process Optimization and Integration
- Venous Thromboembolism Diagnosis and Management
- Innovative Microfluidic and Catalytic Techniques Innovation
- Target Tracking and Data Fusion in Sensor Networks
AstraZeneca (Sweden)
2023-2025
Science for Life Laboratory
2017-2022
KTH Royal Institute of Technology
2017-2022
Deutsches Herzzentrum München
2021
Helmholtz Zentrum München
2021
Biomolecular simulations are intrinsically high dimensional and generate noisy data sets of ever-increasing size. Extracting important features from the is crucial for understanding biophysical properties molecular processes, but remains a big challenge. Machine learning (ML) provides powerful dimensionality reduction tools. However, such methods often criticized as resembling black boxes with limited human-interpretable insight. We use supervised unsupervised ML to efficiently create...
The efficiency of machine learning (ML) models is crucial to minimize inference times and reduce the carbon footprints deployed in production environments. Current employed retrosynthesis generate a synthesis route from target molecule purchasable compounds are prohibitively slow. model operates single-step fashion tree search algorithm by predicting reactant molecules given product as input. In this study, we investigate ability alternative transformer architectures, knowledge distillation...
Calmodulin (CaM) is a calcium sensing protein that regulates the function of large number proteins, thus playing crucial part in many cell signaling pathways. CaM has ability to bind more than 300 different target peptides Ca2+-dependent manner, mainly through exposure hydrophobic residues. How can targets while retaining some selectivity fascinating open question. Here, we explore mechanism selective promiscuity for selected proteins. Analyzing enhanced sampling molecular dynamics...
Synthesis planning of new pharmaceutical compounds is a well-known bottleneck in modern drug design. Template-free methods, such as transformers, have recently been proposed an alternative to template-based methods for single-step retrosynthetic predictions. Here, we trained and evaluated transformer model, called the Chemformer, retrosynthesis predictions within discovery. The proprietary data set used training comprised ∼18 M reactions from literature, patents, electronic lab notebooks....
Calmodulin (CaM) and phosphatidylinositol 4,5-bisphosphate (PIP2) are potent regulators of the voltage-gated potassium channel KCNQ1 (KV7.1), which conducts cardiac IKs current. Although cryo-electron microscopy structures revealed intricate interactions between voltage-sensing domain (VSD), CaM, PIP2, functional consequences these remain unknown. Here, we show that CaM-VSD act as a state-dependent switch to control pore opening. Combined electrophysiology molecular dynamics network analysis...
Free energy landscapes provide insights into conformational ensembles of biomolecules. In order to analyze these and elucidate mechanisms underlying changes, there is a need extract metastable states with limited noise. This has remained formidable task, despite plethora existing clustering methods. We present InfleCS, novel method for extracting well-defined core from free landscapes. The based on Gaussian mixture estimator exploits the shape estimated density landscape. that naturally...
Many membrane proteins are modulated by external stimuli, such as small molecule binding or change in pH, transmembrane voltage, temperature. This modulation typically occurs at sites that structurally distant from the functional site. Revealing communication, known allostery, between these two is key to understanding mechanistic details of proteins. Residue interaction networks isolated commonly used this end. Membrane proteins, however, embedded a lipid bilayer, which may contribute...
Cardiovascular disease (CVD) is the leading cause of morbidity and mortality worldwide. Non-coding RNAs have already been linked to CVD development progression. While microRNAs (miRs) well studied in blood samples, there little data on tissue-specific miRs cardiovascular relevant tissues their relation risk factors. Tissue-specific derived from Arteria mammaria interna (IMA) 192 coronary artery (CAD) patients undergoing bypass grafting (CABG) were analyzed. The aims study 1) establish a...
A free energy landscape estimation method based on the well-known Gaussian mixture model (GMM) is used to compare efficiencies of thermally enhanced sampling methods with respect regular molecular dynamics. The simulations are carried out two binding states calmodulin, and compared other estimators using a toy model. We show that GMM cross-validation provides robust estimate not subject overfitting. continuous nature Gaussians better estimates sparse data than canonical histogramming. find...
The multi-step retrosynthesis problem can be solved by a search algorithm, such as Monte Carlo tree (MCTS). performance of multistep retrosynthesis, measured trade-off in time and route solvability, therefore depends on the hyperparameters algorithm. In this paper, we demonstrated effect three MCTS (number iterations, depth, width) metrics Linear integrated speed-accuracy score (LISAS) Inverse efficiency which consider both solvability time. This exploration was conducted employing...
Designing synthesis routes with shared intermediates for a set of target compounds is common task in drug discovery. Multistep retrosynthesis tools such as AiZynthFinder are frequently used by chemists to generate possible routes. Although these can find solved majority compounds, they may not which comply specific bond constraints. Such constraints could be defined the project obtain compounds. Here, we present novel approach aims feasible region The divided into bonds break and freeze....
KV1.2 channels, encoded by the KCNA2 gene, regulate neuronal excitability conducting K+ upon depolarization. A new missense variant was discovered in a patient with epilepsy, causing amino acid substitution F302L at helix S4, voltage-sensing domain. Immunocytochemistry and flow cytometry showed that does not impair subunit surface trafficking. Molecular dynamics simulations indicated alters exposure of S4 residues to membrane lipids. Voltage clamp fluorometry revealed domain KV1.2-F302L...
Abstract Calmodulin (CaM) is a calcium sensing protein that regulates the function of large number proteins, thus playing crucial part in many cell signaling path- ways. CaM has ability to bind more than 300 different target peptides Ca 2+ -dependent manner, mainly through exposure hydrophobic residues. How can targets while retaining some selectivity fascinating open question. Here, we explore mechanism selective promiscuity for selected proteins. Analyzing enhanced sampling molecular...
ABSTRACT Biomolecular simulations are intrinsically high dimensional and generate noisy datasets of ever increasing size. Extracting important features in the data is crucial for understanding biophysical properties molecular processes, but remains a big challenge. Machine learning (ML) provides powerful dimensionality reduction tools. However, such methods often criticized to resemble black boxes with limited human-interpretable insight. We use from supervised unsupervised ML efficiently...
Synthesis planning of new pharmaceutical compounds is a well-known bottleneck in modern drug design. Template-free methods, such as transformers, have recently been proposed an alternative to template-based methods for single-step retrosynthetic predictions. Here, we trained and evaluated transformer model, called Chemformer, retrosynthesis predictions within discovery. The proprietary dataset used training comprised ~18M reactions from literature, patents, electronic lab notebooks....
We introduce a multi-objective search algorithm for retrosynthesis planning, based on Monte Carlo Tree formalism. The allow combining diverse set of objectives without considering their scale or weighting factors. To benchmark this novel algorithm, we employ four in total eight experiments PaRoutes set. ranges from simple ones starting material and step count to complex synthesis complexity route similarity. show that with the careful employment objectives, outperforms single-objective...
We introduce a multi-objective search algorithm for retrosynthesis planning, based on Monte Carlo Tree formalism. The allow combining diverse set of objectives without considering their scale or weighting factors. To benchmark this novel algorithm, we employ four in total eight experiments PaRoutes set. ranges from simple ones starting material and step count to complex synthesis complexity route similarity. show that with the careful employment objectives, outperforms single-objective...
Efficiency of machine learning (ML) models is crucial to minimize inference times and reduce carbon footprints deployed in production environments. Current employed retrosynthesis generate a synthesis route from target molecule purchasable compounds are prohibitively slow. The model operates single-step fashion tree search algorithm by predicting reactant molecules given product as input. In this study, we investigate the ability alternative transformer architectures, knowledge distillation...
Abstract Many membrane proteins are modulated by external stimuli, such as small molecule binding or change in pH, transmembrane voltage temperature. This modulation typically occurs at sites that structurally distant from the functional site. Revealing communication, known allostery, between these two is key to understanding mechanistic details of proteins. Residue interaction networks isolated commonly used this end. Membrane proteins, however, embedded a lipid bilayer which may contribute...
Abstract Calmodulin (CaM) and PIP 2 are potent regulators of the voltage-gated potassium channel KCNQ1 (K V 7.1), which conducts I Ks current important for repolarization cardiac action potentials. Although cryo-EM structures revealed intricate interactions between voltage-sensing domain (VSD), CaM, , functional consequences these remain unknown. Here, we show that CaM-VSD act as a state-dependent switch to control pore opening. Combined electrophysiology molecular dynamics network analysis...
Calmodulin (CaM) is a calcium sensor which binds and regulates wide range of target-proteins. This implicitly enables the concentration to influence many downstream physiological responses, including muscle contraction, learning depression. The antipsychotic drug trifluoperazine (TFP) known CaM inhibitor. By binding various sites, TFP prevents from associating However, molecular state-dependent mechanisms behind inhibition by drugs such as are largely unknown. Here, we build Markov state...