Yury V. Milchevskiy

Dr. Yury V. Milchevskiy

orcid

Research Focus & Metrics

Academic Career Length

7Years (2020-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
Verified
Journal Article
Open access
PDF file
YMYury V. Milchevskiy
Yury V. Milchevskiy
GKGalina I. Kravatskaya
Galina I. Kravatskaya
YKYuri V. Kravatsky
Yuri V. Kravatsky
The physicochemical properties of amino acid residues from the AAindex database are widely used as predictors in building models for predicting both protein structures and properties. It should be noted, however, that the AAindex database contains data only for the 20 canonical amino acids. Non-canonical amino acids, while less common, are not rare; the Protein Data Bank includes proteins with more than 1000 distinct non-canonical amino acids. In this study, we propose a method to evaluate the physicochemical properties from the AAindex database for non-canonical amino acids and assess the prediction quality. We implemented our method as a bioinformatics tool and estimated the physicochemical properties of non-canonical amino acids from the PDB with the chemical composition presentation using SMILES encoding obtained from the PDBechem databank. The bioinformatics tool and resulting database of the estimated properties are freely available on the author’s website and available for download via GitHub.

20 May 2021 in PLOS ONE
Verified
Journal Article
Open access
PDF file
VMVladislava Milchevskaya
Vladislava Milchevskaya
ANAlexei M. Nikitin
Alexei M. Nikitin
SLSergey A. Lukshin
Sergey A. Lukshin
IFIvan V. Filatov
Ivan V. Filatov
YKYuri V. Kravatsky
Yuri V. Kravatsky
YMYury V. Milchevskiy
Yury V. Milchevskiy
Motivation: Local protein structure is usually described via classifying each peptide to a unique class from a set of pre-defined structures. These classifications may differ in the number of structural classes, the length of peptides, or class attribution criteria. Most methods that predict the local structure of a protein from its sequence first rely on some classification and only then proceed to the 3D conformation assessment. However, most classification methods rely on homologous proteins’ existence, unavoidably lose information by attributing a peptide to a single class or suffer from a suboptimal choice of the representative classes. Results: To alleviate the above challenges, we propose a method that constructs a peptide’s structural representation from the sequence, reflecting its similarity to several basic representative structures. For 5-mer peptides and 16 representative structures, we achieved the Q16 classification accuracy of 67.9%, which is higher than what is currently reported in the literature. Our prediction method does not utilize information about protein homologues but relies only on the amino acids’ physicochemical properties and the resolved structures’ statistics. We also show that the 3D coordinates of a peptide can be uniquely recovered from its structural coordinates, and show the required conditions under various geometric constraints.

Verified
Journal Article
Open access
PDF file
YMYury V. Milchevskiy
Yury V. Milchevskiy
VMVladislava Y. Milchevskaya
Vladislava Y. Milchevskaya
ANAlexei M. Nikitin
Alexei M. Nikitin
YKYuri V. Kravatsky
Yuri V. Kravatsky
Protein structure prediction continues to pose multiple challenges despite outstanding progress that is largely attributable to the use of novel machine learning techniques. One of the widely used representations of local 3D structure—protein blocks (PBs)—can be treated in a similar way to secondary structure classes. Here, we present a new approach for predicting local conformation in terms of PB classes solely from amino acid sequences. We apply the RMSD metric to ensure unambiguous future 3D protein structure recovery. The selection of statistically assessed features is a key component of the proposed method. We suggest that ML input features should be created from the statistically significant predictors that are derived from the amino acids’ physicochemical properties and the resolved structures’ statistics. The statistical significance of the suggested features was assessed using a stepwise regression analysis that permitted the evaluation of the contribution and statistical significance of each predictor. We used the set of 380 statistically significant predictors as a learning model for the regression neural network that was trained using the PISCES30 dataset. When using the same dataset and metrics for benchmarking, our method outperformed all other methods reported in the literature for the CB513 nonredundant dataset (for the PBs, Q16 = 81.01%, and for the DSSP, Q3 = 85.99% and Q8 = 79.35%).

1 February 2023 in Molecular Biology
Verified
Journal Article
Closed access
PDF file
YMYury V. Milchevskiy
Yury V. Milchevskiy
VMV. Yu. Milchevskaya
V. Yu. Milchevskaya
YKYu. V. Kravatsky
Yu. V. Kravatsky
Recently, prediction of the structure and function of a protein from its sequence underwent a rapid increase in performance. It is primarily due to the application of machine learning methods, many of which rely on the predictive features supplied to them. It is thus crucial to retrieve the information encoded in the amino acid sequence of a protein. Here we propose a method to generate a set of complex yet interpretable predictors, which aids in revealing factors that influence protein conformation. The method makes it possible to generate predictive features and test them for significance both in the context of a general description of the protein structures and functions and in the context of highly specific predictive tasks. Having generated an exhaustive set of predictors, we narrow it down to a smaller curated set of informative features using feature selection methods, which increases the performance of subsequent predictive modelling. We illustrate the efficiency of our methodology by applying it to local protein structure prediction, where the rate of correct prediction for DSSP Q3 (three-class classification) is 81.3%. The method is implemented in C++ for command line use and can be run on any operating system. The source code is released on GitHub at https://github.com/Milchevskiy/protein-encoding-projects.

Verified
Journal Article
Open access
PDF file
YMYury V. Milchevskiy
Yury V. Milchevskiy
GKGalina I. Kravatskaya
Galina I. Kravatskaya
YKYuri V. Kravatsky
Yuri V. Kravatsky
Protein structure prediction continues to pose multiple challenges, despite the progress made by ML. While recent deep learning models have achieved a strong performance using embeddings from protein language models, they often ignore non-canonical amino acids and rely heavily on sequence alignments or evolutionary profiles. Here, we present an improvement to this approach for predicting the secondary protein structure of DSSP classes solely from amino acid sequences. We suggest that ML feature sets should be generated from statistically significant mutually uncorrelated descriptors. The selection of statistically assessed descriptors, including predicting the physicochemical parameters of non-canonical amino acids, is a key component of the proposed method. The statistical significance and influence of each of the suggested features were assessed using a two-step Linear Discriminant Analysis, which permitted the evaluation of the statistical significance of each descriptor and their impact on model accuracy. We applied the set of 109 most influential statistically significant descriptors as a learning model for the two-layer Bi-LSTM network combined with ESMFold2 embeddings. Our method, TruMPET (Training upon Multiple Pre-selected Elements Technique), outperformed all other methods reported in the literature for the non-redundant datasets (CB513: DSSP Q3 = 91.36% and Q8 = 85.41%, TEST2018: DSSP Q3 = 90.64% and Q8 = 84.17%).

16 September 2020 in bioRxiv
Verified
Preprint
Open access
PDF file
VMVladislava Milchevskaya
Vladislava Milchevskaya
ANAlexei M. Nikitin
Alexei M. Nikitin
SLSergey A. Lukshin
Sergey A. Lukshin
IFIvan V. Filatov
Ivan V. Filatov
YKYuri V. Kravatsky
Yuri V. Kravatsky
YMYury V. Milchevskiy
Yury V. Milchevskiy
Motivation Local protein structure is usually described via classifying each peptide to a unique element from a set of pre-defined structures. These so-called structural alphabets may differ in the number of structures or the length of peptides. Most methods that predict the local structure of a protein from its sequence rely on this kind of classification. However, since all peptides assigned to the same class are indistinguishable, such an approach may not be sufficient to model protein folding with high accuracy. Results We developed a method that predicts the structural representation of a peptide from its sequence. For 5-mer peptides, we achieved the Q16 classification accuracy of 67.9%, which is higher than what is currently reported in the literature. Importantly, our prediction method does not utilize information about protein homologues but only physicochemical properties of the amino acids and the statistics of the structures, but relies on a comprehensive feature-generation procedure based only on the protein sequence and the statistics of resolved structures. We also show that the 3D coordinates of a peptide can be uniquely recovered from its structural coordinates, and show the required conditions for that under various geometric constraints. Availability The online implementation of the method is provided freely at http://pbpred.eimb.ru Contact milch{at}eimb.ru or vmilchev{at}uni-koeln.de Supplementary information Supplementary data are available online at http://pbpred.eimb.ru/S/index.html