Ivankov Lab
About us
Start date – April 2019
The lab is based at Center of Life Sciences of Skolkovo Institute of Science and Technology. Our main research areas are protein bioinformatics, physics, and evolution.
One research direction of the lab is to improve prediction of protein stability change upon single mutation by combining machine learning and protein physics. Related topics are computational construction of more stable proteins and prediction of protein folding kinetics from protein three-dimensional structure. Another research direction is theoretical investigation of evolution with focus on multi-dimensional epistasis. More specifically, the lab develops computational methods and algorithms to find epistasis in experimental data and to estimate abundance of uni- and multi-dimensional epistasis in protein evolution.
Team
Projects
Prediction of Protein Stability Change Upon Mutation using Deep Learning
Marina Pak
Protein stability is a crucial characteristic of a protein and predicting its alteration upon amino acid substitution is essential for understanding of protein folding mechanisms as well as engineering of proteins with desired properties. Experimental estimation of protein stability change upon mutation, being cumbersome and limitedly executable, makes computational prediction of this parameter a relevant and challenging task.
To date, a great number of computational methods for prediction of protein stability change upon amino acid substitutions have been developed, nevertheless, they are still far from delivering the level of robustness for wide application for protein engineering. Among the most common drawbacks are overfitting, bias towards destabilizing mutations, violation of anti-symmetry principle. The proposed research project is aimed to develop a computational tool for prediction of protein stability change upon mutation combining multiple approaches to overcome the limitations of modern predictors. The predictor is designed to be based on deep neural network trained on a large balanced and symmetrized dataset of experimental data of folding free energy changes upon mutation.
Evolution with focus on multi-dimensional epistasis and evolutionary models
María Carolina Erazo Muñoz
The classical definition of epistasis centers on the interaction between different genes when expressing a certain phenotypic character, i.e. when the expression of one or more genes depends on the expression of others genes. The current understanding of epistasis also includes interaction between positions in protein, RNA, or DNA sequences [1]. The most straightforward approach to detect epistasis is to calculate epistatic coefficients from combinatorially complete datasets, which form hypercubes in sequence space [1, 2]. Most commonly, researchers apply Hadamard-Walsh transformation to get epistatic coefficients [1]; however, the “thermodynamic” transformation could also be applied [2]. Despite the formal link between the two approaches has been established [2], the empirical relationships have not yet been explored.
The aim of the project is to study the correlation between epistatic coefficients obtained by the Hadamard-Walsh and “thermodynamic” approaches using as an example all available fitness landscapes. The combinatorially complete datasets can be found effectively by HypercubeME program [4].
Additionally, we aim at checking which of the evolutionary models is compatible with epistatic pictures observed on the fitness landscapes.
Investigation of epistasis using composite mutations
Evgenii Zorin
Higher order epistasis is calculated by using hypercube structures, however, all the sequences required to create a combinatorially-complete hypercube are not always available. To this end, it is possible to calculate a certain type of epistasis by using composite mutations in hyperrectangles, which act as parallel vectors in the complete multidimensional space of hypercubes. Therefore, the project is to use hypercube-like algorithms to investigate epistasis in hyperrectangles arising from composite mutations.
Publications
Inhibition of hyaluronan secretion by novel coumarin compounds and chitin synthesis inhibitors
Tsitrina AA, Krasylov IV, Maltsev DI, Andreichenko IN, Moskvina VS, Ivankov DN, Bulgakova EV, Nesterchuk M,
Shashkovskaya V, Dashenkova NO, Khilya VP, Mikaelyan A, Kotelevtsev Y.
Glycobiology. 2021, Sep 9;31(8):959-974.
PMID: 33978736
DOI: 10.1093/glycob/cwab038
Solution of Levinthal's Paradox and a Physical Theory of Protein Folding Times
Ivankov DN, Finkelstein AV.
Biomolecules. 2020 Feb 6;10(2):250.
PMID: 32041303
DOI: 10.3390/biom10020250
HypercubeME: two hundred million combinatorially complete datasets from a single experiment
Esteban LA, Lonishin LR, Bobrovskiy D, Leleytner G, Bogatyreva NS, Kondrashov FA, Ivankov DN.
Bioinformatics. 2019 Nov 19;36(6):1960-2.
PMID: 31742320
DOI: 10.3390/biom10020250
An experimental assay of the interactions of amino acids from orthologous sequences shaping a complex fitness landscape
Pokusaeva VO, Usmanova DR, Putintseva EV, Espinar L, Sarkisyan KS, Mishin AS,
Bogatyreva NS, Ivankov DN, Akopyan AV, Avvakumov SY, Povolotskaya IS, Filion GJ, Carey LB, Kondrashov FA.
PLoS Genet. 2019 Apr 10;15(4):e1008079.
PMID: 30969963
DOI: 10.1371/journal.pgen.1008079
Local fitness landscape of the green fluorescent protein
Sarkisyan KS, Bolotin DA, Meer MV, Usmanova DR, Mishin AS, Sharonov GV, Ivankov DN, Bozhanova NG,
Baranov MS, Soylemez O, Bogatyreva NS,
Vlasov PK, Egorov ES, Logacheva MD, Kondrashov AS, Chudakov DM, Putintseva EV, Mamedov IZ, Tawfik DS,
Lukyanov KA, Kondrashov FA.
Nature, 2016, 533: 397‐401.
PMID: 27193686
DOI: 10.1038/nature17995
Thesis defenses
MSc 2020, Marina Pak: Study of influence of homology modeling on the prediction of protein stability change upon mutation
MSc 2021, Natalia Sivitskaia: Search for amino acid substitutions stabilizing human ribonuclease inhibitor