A5070288864- Protein Structure and Dynamics
- Enzyme Structure and Function
- Machine Learning in Materials Science
- Machine Learning in Bioinformatics
- Computational Drug Discovery Methods
- Various Chemistry Research Topics
- Bioinformatics and Genomic Networks
- Graph Theory and Algorithms
- Cardiac electrophysiology and arrhythmias
- Monoclonal and Polyclonal Antibodies Research
- Advanced Graph Neural Networks
- Advanced Neural Network Applications
- Microbial Metabolic Engineering and Bioproduction
- X-ray Diffraction in Crystallography
- Advanced Electron Microscopy Techniques and Applications
- thermodynamics and calorimetric analyses
- Molecular spectroscopy and chirality
- ECG Monitoring and Analysis
- Advanced X-ray Imaging Techniques
- Analytical Chemistry and Chromatography
- Biomedical Text Mining and Ontologies
- RNA and protein synthesis mechanisms
- Gene expression and cancer classification
- Advanced MRI Techniques and Applications
Skolkovo Institute of Science and Technology
2024
Concordia University
2018-2022
CEA Grenoble
2014
Commissariat à l'Énergie Atomique et aux Énergies Alternatives
2014
Institut de Biologie Structurale
2014
Université Grenoble Alpes
2014
Centre National de la Recherche Scientifique
2014
Moscow Institute of Physics and Technology
2014
Forschungszentrum Jülich
2014
Centre Inria de l'Université Grenoble Alpes
2013
Abstract Motivation The computational prediction of a protein structure from its sequence generally relies on method to assess the quality models. Most assessment methods rank candidate models using heavily engineered structural features, defined as complex functions atomic coordinates. However, very few have attempted learn these features directly data. Results We show that deep convolutional networks can be used predict ranking model structures solely basis their raw three-dimensional...
We report the first assessment of blind predictions water positions at protein-protein interfaces, performed as part critical predicted interactions (CAPRI) community-wide experiment. Groups submitting docking for complex DNase domain colicin E2 and Im2 immunity protein (CAPRI Target 47), were invited to predict interfacial molecules using method their choice. The predictions-20 groups submitted a total 195 models-were assessed by measuring recall fraction water-mediated contacts. Of 176...
We study the challenges of applying deep learning to gene expression data. find experimentally that there exists non-linear signal in data, however is it not discovered automatically given noise and low numbers samples used most research. discuss how interaction graphs (same pathway, protein-protein, co-expression, or research paper text association) can be impose a bias on model similar spatial imposed by convolutions an image. explore usage Graph Convolutional Neural Networks coupled with...
Predicting the physical interaction of proteins is a cornerstone problem in computational biology. New classes learning-based algorithms are actively being developed, and typically trained end-to-end on protein complex structures extracted from Protein Data Bank. These training datasets tend to be large difficult use for prototyping and, unlike image or natural language datasets, they not easily interpretable by non-experts. We present Dock2D-IP Dock2DIF, two "toy" that can used select...
Abstract Protein-protein interactions are determined by a number of hard-to-capture features related to shape complementarity, electrostatics, and hydrophobicity. These may be intrinsic the protein or induced presence partner. A conventional approach protein-protein docking consists in engineering small spatial for each protein, minimizing sum their correlations with respect arrangement two proteins. To generalize this approach, we introduce deep neural network architecture that transforms...
HermiteFit , a novel algorithm for fitting protein structure into low-resolution electron-density map, is presented. The accelerates the rotation of Fourier image electron density by using three-dimensional orthogonal Hermite functions. As part new method, an in basis and conversion expansion coefficients are was implemented cross-correlation or Laplacian-filtered as criterion. It demonstrated that Laplacian filter has particularly simple form. To assess quality encoding basis, analytical...
Predicting the structure of a protein from its sequence is cornerstone task molecular biology. Established methods in field, such as homology modeling and fragment assembly, appeared to have reached their limit. However, this year saw emergence promising new approaches: end-to-end dynamics models, well reinforcement learning applied folding. For these approaches be investigated on larger scale, an efficient implementation key computational primitives required. In paper we present library...
Abstract Funding Acknowledgements Type of funding sources: None. Background Deep learning is increasingly used in modern biomedical research and applications due to the substantial availability large clinical datasets. These approaches are invaluable tasks involving noisy imaging data, such as tumour segmentation histological images. In cardiology, a deep approach could be helpful real-time tracking sources arrhythmia, i.e. electrical rotational activity heart. However, existing optical or...
Predicting the physical interaction of proteins is a cornerstone problem in computational biology. New classes learning-based algorithms are actively being developed, and typically trained end-to-end on protein complex structures extracted from Protein Data Bank. These training datasets tend to be large difficult use for prototyping and, unlike image or natural language datasets, they not easily interpretable by non-experts. We present Dock2D-IP Dock2D-IF, two "toy" that can used select...