Showing 1–2 of 2 results for author: Gonczarek, A

Towards fully automated protein structure elucidation with NMR spectroscopy

Authors: Piotr Klukowski, Adam Gonczarek

Abstract: Nuclear magnetic resonance (NMR) spectroscopy is one of the leading techniques for protein studies. The method features a number of properties, allowing to explain macromolecular interactions mechanistically and resolve structures with atomic resolution. However, due to laborious data analysis, a full potential of NMR spectroscopy remains unexploited. Here we present an approach aiming at automati… ▽ More Nuclear magnetic resonance (NMR) spectroscopy is one of the leading techniques for protein studies. The method features a number of properties, allowing to explain macromolecular interactions mechanistically and resolve structures with atomic resolution. However, due to laborious data analysis, a full potential of NMR spectroscopy remains unexploited. Here we present an approach aiming at automation of two major bottlenecks in the analysis pipeline, namely, peak picking and chemical shift assignment. Our approach combines deep learning, non-parametric models and combinatorial optimization, and is able to detect signals of interest in a multidimensional NMR data with high accuracy and match them with atoms in medium-length protein sequences, which is a preliminary step to solve protein spatial structure. △ Less

Submitted 31 July, 2018; originally announced August 2018.

Journal ref: 3rd International Workshop on Biomedical Informatics with Optimization and Machine Learning. ICML 2018

Learning Deep Architectures for Interaction Prediction in Structure-based Virtual Screening

Authors: Adam Gonczarek, Jakub M. Tomczak, Szymon Zaręba, Joanna Kaczmar, Piotr Dąbrowski, Michał J. Walczak

Abstract: We introduce a deep learning architecture for structure-based virtual screening that generates fixed-sized fingerprints of proteins and small molecules by applying learnable atom convolution and softmax operations to each compound separately. These fingerprints are further transformed non-linearly, their inner-product is calculated and used to predict the binding potential. Moreover, we show that… ▽ More We introduce a deep learning architecture for structure-based virtual screening that generates fixed-sized fingerprints of proteins and small molecules by applying learnable atom convolution and softmax operations to each compound separately. These fingerprints are further transformed non-linearly, their inner-product is calculated and used to predict the binding potential. Moreover, we show that widely used benchmark datasets may be insufficient for testing structure-based virtual screening methods that utilize machine learning. Therefore, we introduce a new benchmark dataset, which we constructed based on DUD-E and PDBBind databases. △ Less

Submitted 19 September, 2017; v1 submitted 23 October, 2016; originally announced October 2016.

Comments: Workshop on Machine Learning in Computational Biology. 30th Conference on Neural Information Processing Systems (NIPS 2016), Barcelona, Spain Extended version published in Computers in Biology and Medicine and available online: http://www.sciencedirect.com/science/article/pii/S0010482517302974