Alexey A. Shvets

Crowding on DNA in Protein Search for Targets

Proteins searching and recognizing specific sites on DNA is required for initiating all major biological processes. While the details of the protein search for targets on DNA in purified in vitro systems are reasonably well understood, the situation in real cells is much less clear. The presence of other types of molecules on DNA should prevent reaching the targets, but experiments show that, surprisingly, the molecular crowding on DNA influences the search dynamics much less than expected. We develop a theoretical method that allowed us to clarify the mechanisms of the protein search on DNA in the presence of crowding. It is found that the dimensionality of the search trajectories specifies whether the crowding will affect the target finding. For 3D search pathways it is minimal, while the strongest effect is for 1D search pathways when the crowding particle can block the search. In addition, for 1D search we determined that the critical parameter is a mobility of crowding agents: highly mobile molecules do not affect the search dynamics, while the slow particles can significantly slow down the process. Physical-chemical explanations of the observed phenomena are presented. Our theoretical predictions thus explain the experimental observations, and they are also supported by extensive numerical simulations.

Role of Static and Dynamic Obstacles in the Protein Search for Targets on DNA

Protein search for specific sequences on DNA marks the beginning of major biological processes. Experiments indicate that proteins find and recognize their targets quickly and efficiently. Because of the large number of experimental and theoretical investigations, there is a reasonable understanding of the protein search processes in purified in vitro systems. However, the situation is much more complex in live cells where multiple biochemical and biophysical processes can interfere with the protein search dynamics. In this study, we develop a theoretical method that explores the effect of crowding on DNA chains during the protein search. More specifically, the role of static and dynamic obstacles is investigated. The method employs a discrete-state stochastic framework that accounts for most relevant physical and chemical processes in the system. Our approach also provides an analytical description for all dynamic properties. It is found that the presence of the obstacles can significantly modify the protein search dynamics. This effect depends on the size of the obstacles, on the spatial positions of the target and the obstacles, on the nature of the search regime, and on the dynamic nature of the obstacles. It is argued that the crowding on DNA can accelerate or slow down the protein search dynamics depending on these factors. A comparison with existing experimental and theoretical results is presented. Theoretical results are discussed using simple physical-chemical arguments, and they are also tested with extensive Monte Carlo computer simulations.

Sequence Heterogeneity Accelerates Protein Search for Targets on DNA

The process of protein search for specific binding sites on DNA is fundamentally important since it marks the beginning of all major biological processes. We present a theoretical investigation that probes the role of DNA sequence symmetry, heterogeneity, and chemical composition in the protein search dynamics. Using a discrete-state stochastic approach with a first-passage events analysis, which takes into account the most relevant physical-chemical processes, a full analytical description of the search dynamics is obtained. It is found that, contrary to existing views, the protein search is generally faster on DNA with more heterogeneous sequences. In addition, the search dynamics might be affected by the chemical composition near the target site. The physical origins of these phenomena are discussed. Our results suggest that biological processes might be effectively regulated by modifying chemical composition, symmetry, and heterogeneity of a genome.

Deep Convolutional Neural Networks for Breast Cancer Histology Image Analysis

Breast cancer is one of the main causes of cancer death worldwide. Early diagnostics significantly increases the chances of correct treatment and survival, but this process is tedious and often leads to a disagreement between pathologists. Computer-aided diagnosis systems showed potential for improving the diagnostic accuracy. In this work, we develop the computational approach based on deep convolution neural networks for breast cancer histology image classification. Hematoxylin and eosin stained breast histology microscopy image dataset is provided as a part of the ICIAR 2018 Grand Challenge on Breast Cancer Histology Images. Our approach utilizes several deep neural network architectures and gradient boosted trees classifier. For 4-class classification task, we report 87.2% accuracy. For 2-class classification task to detect carcinomas we report 93.8% accuracy, AUC 97.3%, and sensitivity/specificity 96.5/88.0% at the high-sensitivity operating point. To our knowledge, this approach outperforms other common methods in automated histopathological image classification. The source code for our approach is made publicly available at this https URL

Paediatric Bone Age Assessment Using Deep Convolutional Neural Networks

Skeletal bone age assessment is a common clinical practice to diagnose endocrine and metabolic disorders in child development. In this paper, we describe a deep learning approach to the problem of bone age assessment using data from the 2017 Pediatric Bone Age Challenge organized by the Radiological Society of North America. This dataset consists of 12,600 radiological images. Each radiograph in the dataset is an image of a left hand labeled with bone age and sex of a patient. Our approach introduces a comprehensive preprocessing protocol based on the positive mining technique. We use images of whole hands as well as specific hand parts for both training and prediction. This allows us to measure the importance of specific hand bones for automated bone age analysis. We further evaluate the performance of the suggested methods in the context of skeletal development stages. Our approach outperforms other common methods for bone age assessment.

Automatic Instrument Segmentation in Robot-Assisted Surgery Using Deep Learning.

Semantic segmentation of robotic instruments is an important problem for the robot-assisted surgery. One of the main challenges is to correctly detect an instrument’s position for the tracking and pose estimation in the vicinity of surgical scenes. Accurate pixel-wise instrument segmentation is needed to address this challenge. In this paper we describe our winning solution for MICCAI 2017 Endoscopic Vision SubChallenge: Robotic Instrument Segmentation and its further refinement. Our approach demonstrates an improvement over the state-of-the-art results using several novel deep neural network architectures. It addressed the binary segmentation problem, where every pixel in an image is labeled as an instrument or background from the surgery video feed. In addition, we solve a multi-class segmentation problem, in which we distinguish between different instruments or different parts of an instrument from the background. In this setting, our approach outper-forms other methods in every task subcategory for automatic instrument segmentation thereby providing state-of-the-art results for these problems. The source code for our solution is made publicly available at https://github.com/ternaus/robot-surgery-segmentation

On the Mechanism of Homology Search by RecA Protein Filaments

Genetic stability is a key factor in maintaining, survival, and reproduction of biological cells. It relies on many processes, but one of the most important is a homologous recombination, in which the repair of breaks in double-stranded DNA molecules is taking place with a help of several specific proteins. In bacteria, this task is accomplished by RecA proteins that are active as nucleoprotein filaments formed on single-stranded segments of DNA. A critical step in the homologous recombination is a search for a corresponding homologous region on DNA, which is called a homology search. Recent single-molecule experiments clarified some aspects of this process, but its molecular mechanisms remain not well understood. We developed a quantitative theoretical approach to analyze the homology search. It is based on a discrete-state stochastic model that takes into account the most relevant physical-chemical processes in the system. Using a method of first-passage processes, a full dynamic description of the homology search is presented. It is found that the search dynamics depends on the degree of extension of DNA molecules and on the size of RecA nucleoprotein filaments, in agreement with experimental single-molecule measurements of DNA pairing by RecA proteins. Our theoretical calculations, supported by extensive Monte Carlo computer simulations, provide a molecular description of the mechanisms of the homology search.

How conformational dynamics influences the protein search for targets on DNA

Protein search and association to specific sequences on DNA is a starting point for all fundamental biological processes. It has been intensively studied in recent years by a variety of experimental and theoretical methods. However, many features of these complex biological phenomena are still not resolved at the molecular level. Experiments indicate that proteins can be bound non-specifically to DNA in multiple configurations. But the role of conformational fluctuations in the protein search dynamics remains not well understood. Here we develop a theoretical method to analyze how the conformational transitions affect the process of finding the specific targets on DNA. Our approach is based on discrete-state stochastic calculations that take into account the most relevant physical–chemical processes. This allows us to explicitly evaluate the protein search for the targets on DNA at different conditions. Our calculations suggest that conformational fluctuations might strongly affect the protein search dynamics. We explain how the shift in the conformational equilibrium influences the target search kinetics. Theoretical predictions are supported by Monte Carlo computer simulations.

Angiodysplasia Detection and Localization Using Deep Convolutional Neural Networks.

Accurate detection and localization for angiodysplasia lesions is an important problem in early stage diagnostics of gastrointestinal bleeding and anemia. Gold-standard for angiodysplasia detection and localization is performed using wireless capsule endoscopy. This pill-like device is able to produce thousand of high enough resolution images during one passage through gastrointestinal tract. In this paper we present our winning solution for MICCAI 2017 Endoscopic Vision SubChallenge: Angiodysplasia Detection and Localization its further improvements over the state-of-the-art results using several novel deep neural network architectures. It address the binary segmentation problem, where every pixel in an image is labeled as an angiodysplasia lesions or background. Then, we analyze connected component of each predicted mask. Based on the analysis we developed a classifier that predict angiodysplasia lesions (binary variable) and a detector for their localization (center of a component). In this setting, our approach outperforms other methods in every task subcategory for angiodysplasia detection and localization thereby providing state-of-the-art results for these problems. The source code for our solution is made publicly available at https://github.com/ternaus/angiodysplasia-segmentation

Mechanism of Genome Interrogation: How CRISPR RNA-Guided Cas9 Proteins Locate Specific Targets on DNA

The ability to precisely edit and modify a genome opens endless opportunities to investigate fundamental properties of living systems as well as to advance various medical techniques and bioengineering applications. This possibility is now close to reality due to a recent discovery of the adaptive bacterial immune system, which is based on clustered regularly interspaced short palindromic repeats (CRISPR)-associated proteins (Cas) that utilize RNA to find and cut the double-stranded DNA molecules at specific locations. Here we develop a quantitative theoretical approach to analyze the mechanism of target search on DNA by CRISPR RNA-guided Cas9 proteins, which is followed by a selective cleavage of nucleic acids. It is based on a discrete-state stochastic model that takes into account the most relevant physical-chemical processes in the system. Using a method of first-passage processes, a full dynamic description of the target search is presented. It is found that the location of specific sites on DNA by CRISPR Cas9 proteins is governed by binding first to protospacer adjacent motif sequences on DNA, which is followed by reversible transitions into DNA interrogation states. In addition, the search dynamics is strongly influenced by the off-target cutting. Our theoretical calculations allow us to explain the experimental observations and to give experimentally testable predictions. Thus, the presented theoretical model clarifies some molecular aspects of the genome interrogation by CRISPR RNA-guided Cas9 proteins.