Skip to main content

Dmitri Kireev

Professor

Dmitri Kireev, Ph.D.

Professor, Center for Integrative Chemical Biology and Drug Discovery


dmitri_kireev

PHONE
(919) 843-8457
EMAIL
dmitri.kireev@unc.edu
ADDRESS
Marsico Hall, Room 3205, CB# 7363, Chapel Hill, NC, 27599

ACCEPTING DOCTORAL STUDENTS

Computational Biophysics and Molecular Design (Kireev lab) develops and applies computational tools to advance our understanding of complex biological systems and discover therapies for unmet medical needs.


We are highly active in collaborative translational research. In the Center for Integrative Chemical Biology and Drug Discovery (CICBDD), medicinal chemists, biologists and computational scientists work as a team to discover novel chemical probes and drug leads. We were involved in over 50 translational projects in collaboration with biomedical scientists at CICBDD and other academic labs and small companies. Current and past collaborators include S. Frye and L. James (UNC CICBDD) to develop chemical probes for epigenetic reader proteins [1]; X. Wang (UNC CICBDD), on inhibitors of protein kinases Axl, Mer and FLT3 [2–4]; J. Jin (currently Icahn School of Medicine) on lysine methyltransferases [5]; R. Franzini (U. Utah), on inhibitors of NAD+– dependent enzymes [6]; P. Blancafort (Univ. West. Australia), on peptide modulators of transcription factors EN1/2; L. Parise (UNC School of Medicine and Reveris), on antagonists of Calcium- and Integrin-binding protein 1 (CIB1); R. Hromas (U. Florida), on Metnase inhibitors; G. Johnson and L. Graves (UNC and KinoDyn) [7], on modulation of superenhancers; and with Sirga on small-molecule inhibitors of host tRNA/HIV protein interactions.

The most advanced of the current projects focuses on the discovery of new therapeutics to treat Acute Lymphoblastic Leukemia (ALL) and other cancers. Our project started with a rationale that small-molecule inhibitors of MerTK tyrosine kinase, a proto-oncogene ectopically expressed in childhood ALL, would allow to substantially reduce non-targeted chemotherapeutic regimens. From the very beginning, structure-based design was applied to guide the synthetic effort, which eventually resulted in highly potent and selective Mer kinase inhibitors. The lead compounds have shown excellent survival benefit and drug-like profile for the treatment of leukemia in mouse models. UNC filed multiple patents on these discoveries. Clinical trials started in 2018. The current and future effort is focused on computer-aided design is to further exploit the invented series in order to develop compounds with specific polypharmacology profiles. In particular, in a multi-disciplinary NIH R01-funded project (Wang and Kireev (UNC), Graham (Emory)) dual MerTK/Axl inhibitors are expected to provide novel treatment for cancer by targeting tumor cells and activating anti-tumor immunity.

  1. James LILI, Barsyte-Lovejoy D, Zhong N, Krichevsky L, Korboukh VKVK, Herold JMM, MacNevin CJCJ, Norris JLJL, Sagum CACA, Tempel W, Marcon E, Guo H, Gao C, Huang X-P, Duan S, Emili A, Greenblatt JFJF, Kireev DBDB, Jin J, Janzen WPWP, Brown PJPJ, Bedford MTMT, Arrowsmith CHCH, Frye SVS V. Discovery of a chemical probe for the L3MBTL3 methyllysine reader domain. Nat Chem Biol. 2013;9(3):184–91.
  2. Zhang W, McIver ALL, Stashko MAA, DeRyckere D, Branchford BRR, Hunter D, Kireev D, Miley MJJ, Norris-Drouin J, Stewart WMM, Lee M, Sather S, Zhou Y, Di Paola JAA, Machius M, Janzen WPP, Earp HSS, Graham DKK, Frye SV V, Wang X. Discovery of Mer specific tyrosine kinase inhibitors for the treatment and prevention of thrombosis. J Med Chem. 2013/11/14. 2013;56(23):9693–700.
  3. Minson KA, Smith CC, DeRyckere D, Libbrecht C, Lee-Sherick AB, Huey MG, Lasater EA, Kirkpatrick GD, Stashko MA, Zhang W. The MERTK/FLT3 inhibitor MRX-2843 overcomes resistance-conferring FLT3 mutations in acute myeloid leukemia. JCI insight. 2016;1(3):e85630.
  4. Zhang W, DeRyckere D, Hunter D, Liu J, Stashko MA, Minson KA, Cummings CT, Lee M, Glaros TG, Newton DL, Sather S, Zhang D, Kireev D, Janzen WP, Earp HS, Graham DK, Frye S V, Wang X. UNC2025, a potent and orally bioavailable MER/FLT3 dual inhibitor. J Med Chem. 2014/07/30. 2014;57(16):7031–41.
  5. Vedadi M, Barsyte-Lovejoy D, Liu F, Rival-Gervier S, Allali-Hassani A, Labrie V, Wigle TJJ, Dimaggio PAA, Wasney GAA, Siarheyeva A, Dong A, Tempel W, Wang S-CC, Chen X, Chau I, Mangano TJJ, Huang X-PP, Simpson CDD, Pattenden SGG, Norris JLL, Kireev DBB, Tripathy A, Edwards A, Roth BLL, Janzen WPP, Garcia BAA, Petronis A, Ellis J, Brown PJJ, Frye SV V, Arrowsmith CHH, Jin J. A chemical probe selectively inhibits G9a and GLP methyltransferase activity in cells. Nat Chem Biol. 2011/07/12. 2011;7(8):566–74.
  6. Yuen LH, Dana S, Liu Y, Bloom SI, Thorsell A-G, Neri D, Donato AJ, Kireev D, Schüler H, Franzini RM. A Focused DNA-Encoded Chemical Library for the Discovery of Inhibitors of NAD+-Dependent Enzymes. J Am Chem Soc. 2019 Apr 3;141(13):5169–81.
  7. Krulikas LJ, McDonald IM, Lee B, Okumu DO, East MP, Gilbert TSK, Herring LE, Golitz BT, Wells CI, Axtman AD, Zuercher WJ, Willson TM, Kireev D, Yeh JJ, Johnson GL, Baines AT, Graves LM. Application of Integrated Drug Screening/Kinome Analysis to Identify Inhibitors of Gemcitabine-Resistant Pancreatic Cancer Cell Growth. SLAS Discov Adv Life Sci R&D. 2018 Sep 9;23(8):850–61.


Dr. Kireev’s early research focused on Quantitative Structure-Activity Relationships (QSAR) and Machine Learning. By then, QSAR had proved a useful strategy in pharmaceutical research. However, the basic QSAR techniques (invented back in 1960’s and 70’s) were seriously challenged by the advent of high-throughput discovery technologies that led to an explosion of chemical, biological and structural information. To address this challenge, I joined an effort of adopting and developing more performant machine learning techniques for chemoinformatics and QSAR. In particular, Dr. Kireev developed and published a new neural network approach, ChemNet [8], capable of handling an unlimited amount of chemical and biological information. He was also among the first to apply self-organizing neural networks – a technique emerged from the Artificial Intelligence (AI) research – to visualization of large quantities of chemical and biological information.

At UNC-CH, Kireev lab developed new CADD for hit and lead identification. One of them, Structural Protein-Ligand Interaction Fingerprints (SPLIF) [9], enables a quantitative score of whether a docking pose interacts with the protein target similarly to a known ligand, which helps to improve both sensitivity and specificity of Structure-Based Virtual Screening. A group at Stanford included SPLIF in their DeepChem (deepchem.io) suite, a toolkit of choice for the developers of AI applications in chemistry. Most recently, we developed a novel approach for data-driven rational design making use of FRAgments in Structural Environments (FRASE) [10]. FRASE-based design distills 3D information relevant to the protein of interest from tens of thousands 3D structures in the Protein Data Bank (PDB) and of millions of structure-activity relationships (SAR) from multiple online sources. Remarkably, unlike most other published computational approaches that are rather ‘virtual screeners’ (that is, predict potencies for existing chemical compounds), ours is a ‘virtual medicinal chemist that designs yet nonexistent compounds. Eventually, FRASEs can be used as components for building multi-target inhibitors with a specific selectivity profile.

Our current effort is focused on the development of FRASE-bot, a fully automated artificial intelligence (AI)-based system to design ligands for promising ligand-orphan targets. FRASE-bot will overcome the major limitation of the previously published FRASE-based approach. The latter only allows to exploit PDB/SAR information within a given protein family, hence precluding the ligand discovery for the majority of novel targets of interest belonging to understudied families. In contrast, FRASE-bot will allow exploiting the full body of 3D structural and SAR data to assemble a ligand in the binding site of any orphan protein. Moreover, FRASE-bot includes an AI-based generative component that is able to automatically construct valid 3D structures of small-molecule ligands from FRASEs directly in the protein pocket. We expect this project to eventually grow into a startup company offering ligand discovery services to pharmaceutical companies.

Other ongoing and future projects include applications of deep learning to improve the outcomes of DNA-encoded libraries, in collaboration with Pearce and Franzini groups (respectively, UNC and UUtah), and use of AI on vast amounts of SAR, genomic, in vitro and in vivo data to better predict the effect of small-molecule drugs in animal models and human patients.

  1. Kireev D. ChemNet: A Novel Neural Network Based Method for Graph/Property Mapping. J Chem Inf Comput Sci. 1995;35(2):175–80.
  2. Da C, Kireev D. Structural Protein–Ligand Interaction Fingerprints (SPLIF) for Structure-Based Virtual Screening: Method and Benchmark Study. J Chem Inf Model. 2014;54(9):2555–61.
  3. 10. Da C, Zhang D, Stashko M, Vasileiadi E, Parker RE, Minson KA, Huey MG, Huelse JM, Hunter D, Gilbert TSK, Norris-Drouin J, Miley M, Herring LE, Graves LM, DeRyckere D, Earp HS, Graham DK, Frye S V., Wang X, Kireev D. Data-Driven Construction of Antitumor Agents with Controlled Polypharmacology. J Am Chem Soc. 2019 Oct 2;141(39):15700–9.


Another important part of research in Kireev lab involves multi-scale simulations of biological systems. All-atom molecular dynamics (MD) can be successfully exploited to better understand structural mechanisms underlying the function of therapeutically important proteins or to develop chemical probes for challenging protein targets. The following published studies showcase usefulness of this approach:

  • In an early study, alchemical free energy simulations were used to understand physical principles of how epigenetic “reader” proteins are able to distinguish between subtly different lysine methylation states of the histone H3 peptide [11].
  • Later, a multidisciplinary collaboration with the Strahl and Arrowsmith groups (respectively, UNC and UToronto) resulted in obtaining a unique insight into the structural mechanism by which the histone H3 engages the Tandem Tudor and PHD fragments (TTD-PHD) of UHRF1, an E3 ubiquitin-protein ligase overexpressed in many human cancers. A series of microsecond-scale non-equilibrium molecular dynamics simulations for the UHRF1 TTD–PHD has been performed to recover a collective average of the free-energy landscape. The landscape analysis allowed us to predict the sequence of events (an entropic lock-and-key mechanism) leading to an effective bivalent engagement of respectively TTD and PHD to two distinct sites on the H3 peptide [12].
  • More recently we collaborated with Shears group (NIEHS) to rationalize the catalytic versatility of PPIP5K2, an enzyme with critical roles in cell signaling and bioenergetic homeostasis. Microsecond-scale all-atom MD simulations revealed a distinctive ratchet mechanism involved in intramolecular substrate transfer. Moreover, a new substrate was computationally predicted, synthesized and experimentally confirmed [13].
  • Another recent example involves the application of MD simulations to the design of allosteric mutant-specific activators of Polycomb repressive complex 2 (PRC2) that regulates gene activity by trimethylation of histone H3 lysine 27. Computational simulations allowed a better understanding of how small-molecule ligands can be turned either into PRC2 agonists or antagonists by stabilizing an a-helix of EZH2, a methyltransferase, adjacent to the EED subunit. Furthermore, MD simulations revealed the structural mechanism by which the EED-I363M mutation inhibits the catalytic activity of EZH2. The simulation model was then used to design ligands that would preferentially bind to the mutant EED subunit and correct its loss-of-function effect by interacting EZH2.

To simulate biological systems at a larger scale, we are developing Molecular Biosystems (MB), an ultra-coarse-grained approach to enable investigation of physical principles underlying behavior of heterogeneous molecular ensembles. MB represents a blend of top-down and bottom-up coarse-grained models. It makes use of abundant experimental data on protein-protein affinities and enzyme kinetics to parameterize stochastic processes enabling formation of protein complexes and enzyme-mediated transformations. On the other hand, motions of MB particles in cytosol can be simulated from physical principles using Brownian dynamics. As a first model system for MB [14], we selected regulatory regions in the Oct4 locus, a key gene controlling gene pluripotency, highly significant in regenerative medicine and cancer research. Surprisingly, such a simple unbiased model provides a comprehensive structural interpretation of micro- and macroscopic heterochromatin properties. In particular, it sheds light on the structural roots of a yet poorly understood phenomenon of a non-deterministic nature of heterochromatin formation and subsequent gene repression. The simulations revealed how an extremely low rate of microscopic heterochromatin-driving events – formation of quaternary H3-HP1 complexes – makes the fiber condensation (associated with repressed chromatin state) a non-deterministic process. Our collaboration with N. Hathaway (UNC) enabled experimental corroboration of the computational outcomes[14].

In the future, Molecular Biosystems will be applied to increasingly large systems up to a whole cell. This expansion will require further developments to significantly enhance the software performance and to design representation of yet unaccounted biological objects, such as organelles.

Other future plans involve application of the MB platform to simulations of simple, but biologically accurate, particle-based 3D models of neuronal networks. These models will help to better understand physical principles of how memory and cognition emerge and evolve from competitive interactions of interconnected neurons to external stimuli. This research will help to develop more compact and energy-efficient AI systems.

  1. Gao C, Herold JM, Kireev D, Wigle T, Norris JL, Frye S. Biophysical Probes Reveal a “Compromise” Nature of the Methyl-lysine Binding Pocket in L3MBTL1. J Am Chem Soc. 2011;133(14):5357–62.
  2. Rothbart SBSB, Dickson BMBM, Ong MSMS, Krajewski K, Houliston S, Kireev DBDB, Arrowsmith CHCH, Strahl BDBD. Multivalent histone engagement by the linked tandem Tudor and PHD domains of UHRF1 is required for the epigenetic inheritance of DNA methylation. Genes Dev. 2013;27(11):1288–98.
  3. An Y, Jessen HJ, Wang H, Shears SB, Kireev D. Dynamics of Substrate Processing by PPIP5K2, a Versatile Catalytic Machine. Structure. 2019;27(6):1022–8.
  4. Williams MR, Xiaokang Y, Hathaway NA, Kireev D. Submolecular-resolution 3D Simulations of the Oct4 Promoter Region Predict Structural Mechanism of Heterochromatin Formation. Submitted.


 


Dmitri Kireev News