I will cover machine learning methodologies that my lab has developed for the design and engineering of protein therapeutics, including antibodies and Zinc Finger proteins. Our methods integrate modern deep learning methods, classical protein modeling and high-throughput experiments.

The behaviors of biomolecules at aqueous-nonaqueous interfaces markedly differ from those occurring in bulk water. First, we found that various biomolecular reactions are highly accelerated by a factor of 103 or higher at the aqueous interface of micron-sized water droplets (microdroplets). Second, thermodynamically unfavorable biomolecular reactions such as phosphorylation and RNA synthesis can proceed in water microdroplets at room temperature without any added energy sources such as ATP. We found that molecules are in highly ordered states near the aqueous-nonaqueous interface, which provides an environment for minimized entropic barriers for biomolecular reactions. Third, the charge exchange between water molecules and biomolecules spontaneously occurs at the water interface, leading to the oxidation of water molecules and the reduction of biomolecules. Vibrational Stark spectroscopy coupled with stimulated Raman excited fluorescence microscopy revealed the presence of a strong electric field formed at the aqueous-nonaqueous interface. This interfacial electric field combined with the partial desolvation effect can be in part responsible for the observed behaviors. Other empirical and theoretical investigations of microdroplet chemistry and their biological implications will be discussed.

대한민국 혁신신약 연구의 시작은 GATT* 도입을 위한 1987년도 물질특허의 도입부터라고 생각합니다. 물질특허가 도입되면서 신물질 복제가 불가능해짐에 따라 당시 자본력이 있던 재벌계열 제약회사들(LG, SK)와 선두급 제약회사 (동야제약, 유한양행)들이 새로운 물질 합성을 통한 신약발굴에 임하였습니다. 첫 성과물들이 1997년도 LG, SK, 유한양행이 당시 글로벌 제약사들에게 Licensing-out을 하면서 나왔습니다. 여기에 휴먼지놈프로젝트/ IT 창업열풍과 함께 2000년도 국내에서도 벤처붐이 일어나면서 많은 대학교수, 연구기관 연구자들이 창업열기에 뛰어들었습니다. 이러한 바이오 창업 붐은 뒤이은 불경기로 오랜 시련기를 겪었습니다만, 그 기간 동안 대한민국의 혁신신약을 위한 “기초체력”들은 차근차근 성장을 하였고, 2015년 한미약품의 연속적인 글로벌 라이센싱아웃은 새로운 열기를 불러일으켰습니다. 2016년도부터 벤처창업과 투자, 그리고 성장한 기초과학 실력 등이 합쳐지면서 바이오 창업 붐이 다시 일어서 2021년도까지 이어졌고, 2022년부터는 다시 혹한기를 겪고 있습니다. 이러한 한국 혁신신약의 30여년간을 함께 하면서 저도 약 8년간의 대기업 연구원 및 사업개발 중간관리자 역할과 20여년간의 벤처 경험 (3번의 창업)을 하면서 그 길을 함께 해 왔습니다. 저는 2020년대 후반과 30년대에 한국의 제약바이오 산업이 글로벌 플레이어로 성장할 것을 확신합니다. 이번 발표를 통해서 제가 이렇게 확신을 가지게 된 배경을 여러분들과 공유하고 싶습니다.

We consider quantum dynamics on a graph, with repeated strong measurements performed locally at a fixed time interval τ. For example a particle starting on node x and measurements performed on another nodex′. From the basic postulates of quantum mechanics the string of measurements yields a sequence no,no,no, and finally in the n-th attempt a yes, i.e. the particle is detected, and a hitting event is recorded. Statistics of the first detection time nτ are investigated, and compared with the corresponding classical first passage problem. Dark states, Zeno physics, a quantum renewal equation, winding number for the first return problem (work of A. Grunbaum et al.), total detection probability, detection time operators and time wave functions are discussed. For future practical applications we discuss efficient quantum search on designed special graphs and simple experiments on quantum computers.

Much of our understanding of protein function and mechanism of action is deduced from a static 3D structure of a native conformational state of the protein. However, proteins are soft, dynamic molecules that constantly move along complex free-energy landscapes, either through equilibrium fluctuations or through the input and dissipation of energy. Characterizing the free-energy landscapes and the mechanism of sampling alternative states remains an outstanding challenge in biophysics. In the first part of the talk, I will discuss our studies of equilibrium reaction dynamics occurring along an internal reaction coordinate of green fluorescent protein. It is well known that the chromophore pocket of GFP and its variants show both ground state and excited state conformational exchange. Using single–molecule spectroscopy, we demonstrate that ms equilibrium exchange between two ground state conformations displays a complex coupling to the solvent, consistent with internal friction observed in protein folding experiments. Secondly, I will briefly discuss our studies on biological condensation. Since the widespread observation of liquid–liquid phase separation (LLPS) by proteins rich in intrinsic disorder both in vitro and in vivo, there has been a concerted effort to deduce a structure–function relationship that would allow LLPS to be predicted on the basis of protein structure. We propose a quantitative framework based on protein unfolding thermodynamics capable of predicting both formation and stability of biological condensates.

From Fizeau’s experiment in 1851 on the speed of light to the most recent detection of gravitational waves by the LIGO observatory, optical interferometers have consistently redefined what constitutes the limits of metrology. In this talk, we will explore innovative scattering-based interferometric techniques that came out only less than 5 years ago, and which can effectively bring this powerful tool to probe single molecule systems and highly diluted suspensions: First, Cavity-Amplified Scattering Spectroscopy (CASS) which fundamentally elongates the effective optical path length of scattering suspensions by a factor 102 to 103 and considerably increases the sensitivity limit of conventional scattering techniques by a factor going up to 104-fold. This technique opens the way for DLS-like nanoparticle sizing for highly transparent and miniaturized samples. Then, the recently developed Mass Photometry, which allows for the marker-free quantification of the mass of single molecules through their scattering signal only. Be it protein oligomerization, protein-protein interaction of macromolecular complex assembly, this method makes use of an arsenal of clever techniques to extract this barely detectable signal from the noisy, speckle-ridden images it comes from. Here, we will explore those techniques and ponder upon the future of optical interferometry in this context.

Phase separation involving biomolecular liquids is emerging as a fundamental mechanism of intracellular organization. Liquid-liquid phase separation (LLPS) of biomolecules compartmentalizes cellular contents into multiple coexisting condensates to facilitate spatio-temporal regulation of biological processes. Consistent with their critical roles in cell physiology, aberrant forms of condensates are found in various diseases such as cancers and neurodegeneration. In this talk, I will first discuss optogenetic-based technologies enabling the dynamic control of phase separation in living cells. Using these technologies, we demonstrate that intracellular phase space can be navigated at high spatiotemporal resolution, revealing different material states accessible in cells as well as functional consequences of biomolecular condensation. I will then move on to our recent efforts to understand biophysical principles of intracellular phase separation. Using refractive index imaging, we find that a class of condensates has surprisingly low biomolecular densities. RNA plays a critical role on lowering the density of condensates. The way condensates nucleate and grow is heavily influenced by surrounding intracellular structures, which highlights the role of crowded cellular environments on phase separation. I will end my talk by briefly discussing how the principle of biomolecular phase separation can be utilized to build a novel synthetic system with programmable composition and functionalities.

Recent advances in computational hardware and software have led to the widespread use of first-principle quantum mechanics methods for precise electronic structure calculations and molecular mechanics methods for simulating realistic dynamic behavior of molecular systems. The multi-scale simulation framework, which combines both quantum and molecular mechanics approaches, is particularly promising for analyzing complex solid-liquid interface systems. In this presentation, we will introduce a novel multi-scale simulation methodology, “DFT-CES (Density Functional Theory in Classical Explicit Solvents),” along with its application examples. We will provide an accessible introduction to computational chemistry simulation frameworks and discuss various energy and environmental applications that employ atomic-level multi-scale simulation approaches, including CO2 conversion, fuel cell catalyst systems, and more. We will also present research findings on the direct observation and quantitative thermodynamic analysis of complex electrochemical double layer (EDL) phenomena, which offer valuable insights into the underlying physics of solid-liquid interface interactions, using the DFT-CES method.

The movement of charges at the interface between molecules and electrodes is of significant interest to various research fields, including electronics, photonics, electrosynthesis, energy conversion, sensing, and spectroscopy. However, studying this movement can be challenging since most devices and systems contain poorly defined organic-electrode interfaces, making it difficult to reliably distinguish one effect from another. This presentation describes our recent efforts to investigate the fundamental mechanisms of charge transport in well-defined interfaces created by self-assembled monolayers and electrodes. We incorporated various chemical structures into the monolayers and used the liquid metal technique to measure the rate of charge transport by tunneling across a metal/monolayer/metal oxide/metal junction, in the presence of an external bias voltage or temperature gradient. Through our experiments, we discovered a collection of interesting structure-transport relationships that highlight the interplay between (supra)molecular chemistry, tunneling, and thermoelectricity. These findings provide valuable insights into the behavior of charge movement at interfaces, and could potentially inform the development of new technologies in a range of fields.

The theoretical description of the diffusion-influenced reaction rates has usually been based on the Smoluchowski diffusion equation.1-6 This description captures explicitly two essential components of the reaction process: the diffusive motion towards the reaction surface and finite intrinsic reactivity on it. It has provided an adequate explanation of many experimental observations and computer simulation results, but it fails to give an exact account of the shorttime scale reaction dynamics.7 In the Smoluchowski description, the value of the classical diffusive flux at a given spatial point at a given time is assumed to depend only on the values at the same point and the same moment. In reality, however, relaxation processes have a certain finite duration, i.e., they are inertial. The validity of the diffusion-equation description has been assessed by molecular dynamics simulations8-11 or Langevin dynamics simulations (that is, Brownian dynamics simulations with inertial effects included).12,13 The molecular dynamics simulations revealed the effects of nondiffusive dynamic behavior in two aspects. The first is the non-Markovian dynamic effect arising from finite relaxation times of the solvent motions, and the second is the inertial dynamic effect (also called the memory effect3) arising from the finite velocity relaxation time of the reactant molecules. Therefore, to assess the inertial dynamic effect separately, the use of Langevin dynamics simulations is more appropriate. Analytic theories which dealt with the inertial dynamic effect were proposed by using the Fokker-Planck-Klein-Kramers equation (FPKKE). Unfortunately, analytical solutions to the FPKKE are available only for a few simple cases.14-20In the present work, we will take into account inertial dynamic effects by developing a theory based on the linear diffusive Cattaneo system21-25 with the reaction sink term. Previous analytical studies on the inertial dynamic effects were limited to the bulk recombination reaction with infinite intrinsic reactivity. In the present work, we investigate the combined effects of the inertial dynamics and the finite reactivity on both bulk and geminate recombination rates. We obtain formally exact expressions for the kinetics of both bulk and geminate recombination. The rate expressions reveal the role of inertial dynamic effects at short times, which retard the reaction rates. The time-dependent rate expression for the bulk recombination reactions obtained in the present theory agree with the rate expression of Harris16 that was obtained from an approximate solution of the FPKKE, when the equilibrium rate constant is set equal to that expected from the kinetic theory of dilute gases. On the other hand, when the reactivity at contact distance becomes infinity, our rate expression agrees with that obtained by Rice [Eq. (289) of Ref. 3]. To the knowledge of the authors, the inertial effect on the geminate recombination reaction is first investigated in the present work. In particular, we find a distinctive feature of the inertial dynamic effect on the survival probability of a geminate pair at short times, which can be manifested in experimental observations.

In metazoan, the presence of double-stranded (ds)DNA in the cytosol signals grave issues such as exposure to ionizing irradiation and replicating viruses. The innate immune system acts as the first line of defense against cytosolic dsDNA by initiating inflammatory responses. Cytosolic dsDNA sensing pathways are integral to host defense against numerous pathogens, yet their dysregulation results in various autoimmune disorders and cancer. Here, we delineate the activation and regulation mechanisms of three major cytosolic dsDNA sensors, namely cyclic-G/AMP synthase (cGAS), absent-in-melanoma-2 (AIM2), and interferon-inducible protein 16 (IFI16). Upon dimerizing on dsDNA, cGAS generates cyclic-G/AMP (cGAMP), a key second messenger for activating type-I interferons. Using X-ray crystallography and biochemical measurements, we show here how ligand-mediated allostery regulates the activation of cGAS. AIM2 and IFI16 assemble filamentous signaling platforms on dsDNA, leading to inflammasome formation. Using cryo-EM, in silico simulation, and biochemical assays, we show that filament assembly underpins the nucleic acid specificity of both sensors. Moreover, we find that variable axial and lateral interactions within homologous helical assemblies underpin their signaling direction and specificity.

The convergence between accurate quantum-mechanical (QM) models (and codes) with efficient machine learning (ML) methods seem to promise a paradigm shift in molecular simulations. Many challenging applications are now being tackled by increasingly powerful QM/ML methodologies. These include modeling covalent materials, molecules, molecular crystals, surfaces, and even whole proteins in explicit water (https://arxiv.org/abs/2205.08306). In this talk, I attempt to provide a reality check on these recent advances and on the developments required to enable fully quantum dynamics of complex functional (bio)molecular systems. Multiple challenges are highlighted that should keep theorists in business for the foreseeable future: (1) Ensuring the accuracy of high-level QM methods (https://doi.org/10.1038/s41467-021-24119-3), (2) Describing intricate QM long-range interactions (https://doi.org/10.1126/sciadv.aax0024; https://doi.org/10.1126/science.aae0509; https://doi.org/10.1103/PhysRevLett.128.106101), (3) Treating quantum electrodynamic effects that become relevant for complex molecules (https://doi.org/10.1021/acs.jpclett.1c04222; https://doi.org/10.1103/PhysRevResearch.4.013011). (4) Developing increasingly accurate, efficient, scalable, and transferable ML architectures for molecules and materials (https://doi.org/10.1038/s41467-022-31093-x; https://arxiv.org/abs/2209.14865; https://arxiv.org/abs/2209.03985). (5) Accounting for the quantum nature of the nuclei and the influence of external environments (https://doi.org/10.1038/s41467- 020-20212-1; https://doi.org/10.1038/s41467-022-28461-y). I argue that only a conjunction of all these developments will enable the long-held dream of fully quantum (bio)molecular simulations.

Many polyketide antibiotics are synthesized by multi-enzyme systems that operate in an assembly line-like fashion. Their apparently modular architecture has opened the door to engineering of new antibiotics by rationally manipulating the DNA encoding these megasynthases. Elucidating the structures and mechanisms of these assembly lines represents a challenging frontier in physical and chemical enzymology. While much progress has been made on some fronts over the past three decades, our understanding of other fundamental features of these enzymatic assembly lines is rudimentary. My lecture will provide an overview of what we know and what we don’t know, with a focus on the 6-deoxyerythronolide B synthase, a prototypical assembly-line polyketide synthase. In particular, I will emphasize the opportunities for computational methods to advance our understanding of the mechanistic principles of assembly-line polyketide biosynthesis.

The origin of the synergistic catalytic effect between metal catalysts and reducible oxide has been debated for decades. Clarification of this effect, namely the strong metal–support interaction (SMSI), requires an understanding of the geometric and electronic structures of metal–metal oxide interfaces under operando conditions. A bimetallic Pt alloy or mixed catalyst is an excellent platform to uncover the contentious role of the metal–metal oxide interface because the alloyed transition metal can coexist with the Pt surface layer in the form of an oxidized species on the bimetal surface during catalytic reactions. In this talk, I present in-situ observation results of structural modulation on Pt-based bimetal catalysts and mixed catalysts and its impact on the catalytic activity. We utilized PtNi, and PtCo that includes both of single crystal and nanoparticle surfaces as model catalysts, and carried out various operando surface techniques including ambient pressure scanning tunneling microscopy, ambient pressure X-ray photoelectron spectroscopy, and environmental transmission electron microscopy. We show the coexistence of Pt and metal oxide leads to the enhancement of catalytic activity, indicating these metal-oxide interfaces provide more efficient reaction path for CO oxidation. The mixed catalysts composed of Pt nanoparticles and the mesoporous cobalt oxide exhibit the enhancement of catalytic activity while Pt is encapsulated by the oxide thin layers forming the reactive metal-oxide interfaces. Our results show that the presence of metal-oxide interfaces and the charge transfer through the interface have significant implications in the surface reaction of heterogeneous catalysis.

Perylene diimides (PDIs) are versatile class of organic compounds with unique optical, electronic, and structural properties that make them promising materials for various optoelectronic application. In this talk, I will introduce the studies investigating three perylene diimide (PDI) core dyes with different side-chains at the imide positions self-assembled into aggregates in solution, atop a substrate, and via a combination of these approaches that encourages seeded growth, all using a mixture of two organic solvents. We analyzed the morphology, molecular organization, and emission properties within the aggregates. Even though the PDI assemblies appeared structurally similar, their photophysical behavior was distinct, indicating differences in molecular arrangement. From spectroscopic measurements supported by theoretical calculations, we determined that PDI molecules with alkyl chains at the imide position mainly produced H-dominant aggregates with unidirectional intermolecular orientation, while PDI molecules with short branched hydrocarbon substituents at the imide positions produced J-dominant aggregates with alternating orientations between adjacent PDI molecules. We also noted exceptional cases where the combination of fabrication methods and control over external factors can result in various types of aggregates despite all originating from a single type of molecular structure. These results offer insights into the structure-property correlations of PDI aggregates with different side chains and their potential applications in optoelectronic devices and sensors.

A crucial step in the aggregation of amyloidogenic sequences is the transition from a relatively disordered state to assembly-competent structures referred to as N* states. The assembly-competent N* states share key structural characteristics with the protein monomer in the fibril environment as observed in experiments. Using a coarse-grained model for intrinsically disordered proteins (IDPs), we demonstrate that the N* states exist as rare excited states on the monomer free energy landscape. By relating the populations of the N* states to fibril formation time scales, using an empirical relation derived from lattice simulations, we quantitatively correlate the probability of observing the N* states with the relative aggregation propensities. Using kinetic transition networks, we observe that the N* states are obligatory along the favored assembly routes. I will discuss recent applications of N* theory to explore the origins of polymorphism in fibrils composed of amyloid-β (Aβ) peptide, serum amyloid A (SAA) protein, and fused-in-sarcoma (FUS) protein.

Electrochemistry has become a key player in establishing a global sustainable energy landscape. Unfortunately, most electrochemical processes are limited in their efficiency and selectivity which has prevented them from replacing carbon-intensive industrial processes. Computational simulations can help on various sides to improve electrochemistry, from the fundamental, atomistic understanding, over multi-scale modeling of realistic devices up to high-throughput screening of electrocatalysts. In this talk, I want to summarize our group’s efforts on these three points. Hereby, I will, in particular, speak about the electrochemical reduction of CO2 which has shown its promise in reducing greenhouse gas emissions and at the same time generating a sustainable carbon feedstock. I will show that in this context the electric double layer well-known from electrochemistry is an essential part of the story, affecting all three areas of computational electrochemistry.

Neurons communicate other neurons and other types of cells by releasing neurotransmitters from synaptic vesicles in presynaptic terminals. Recent studies have shown that the spontaneous (i.e., activity-independent) release of neurotransmitters plays an important role in neuronal function and in synaptic development. Moreover, impaired spontaneous release of synaptic vesicles has been implicated in several neurodevelopmental and psychiatric disorders. However, our understanding of spontaneous synaptic vesicle release is quite limited. Specifically, the dynamics and exocytosis of single inhibitory synaptic vesicles undergoing spontaneous release has not been investigated, even though spontaneous release occurs frequently in mature inhibitory synapses and given that inhibitory synaptic transmission plays a critical role in maintaining the delicate balance between neuronal excitation and inhibition. Importantly, changes in this balance have been shown to underlie several neurological disorders, including autism spectrum disorder, schizophrenia, and epilepsy. Using advanced real-time three-dimensional nanometer-accuracy tracking microscope, which allows us to track single synaptic vesicles loaded with a quantum dot (QD) with an accuracy of nanometers in three dimensions, we have studied the dynamics and exocytosis of single spontaneously recycled inhibitory synaptic vesicles in living neurons. We found that the dynamics of spontaneously recycled inhibitory synaptic vesicles undergoing spontaneous release are different from depolarization-evoked recycled inhibitory synaptic vesicles. We also found that a significant number of spontaneously recycled synaptic vesicles undergo exocytosis during the application of external electrical stimuli. Furthermore, these synaptic vesicles exhibit different dynamics compared with synaptic vesicles undergoing spontaneous release. Thus, our results suggest that spontaneously inhibitory recycled synaptic vesicles consist of two distinct subgroups of inhibitory synaptic vesicles and these distinct subgroups of synaptic vesicles contribute to inhibitory synaptic transmission differently. Thus, our results provide new insights into inhibitory synaptic transmission in living neurons. In addition, we tracked single GABAergic synaptic vesicles and GABAergic receptors in living neurons simultaneously during electrical stimuli. The motion of these GABAergic synaptic vesicles and receptors was heterogenous depending on locations. Some GABAergic vesicles and receptors moved juxtaposed across synapses, which indicates dynamic nanocolumn for synaptic transmission. Thus, real-time tracking of single synaptic vesicles and receptors offers new insights to synaptic transmission in living neurons.

Cells at the site of an infection experience numerous biochemical signals that vary in amplitude, space, and time. Despite the diversity of dynamic signals produced by pathogens and sentinel cells, information-processing pathways converge on a limited number of central signaling nodes to ultimately control cellular responses. In particular, the NF-κB pathway responds to dozens of signals from pathogens and self, and plays a vital role in processing proinflammatory inputs. We employed high-throughput microfluidics to investigate the regulation of the NF-κB network in diverse spatial and temporal contexts; then, we analyzed single-cell responses using various mathematical methods, including ODE-based modeling, information theory, and machine/deep learning. Our findings indicate that NF-κB responds sensitively to changes in the level of signal molecules, rather than the magnitude of the level itself. Moreover, experiments involving different sequences of signals revealed that the NF-κB network retains memories of previous signals, influencing its subsequent responses. In addition, our co-culture experiments simulating the tissue environment demonstrated that these dynamics-sensing capabilities of NF-κB enable cells to capture detailed information about the spatial distance and composition of cell types within the tissue. Collectively, our results indicate that the NF-κB network encodes signal dynamics, allowing it to interpret various spatial and temporal changes in the environment and regulate immune responses accordingly.

The talk examines the influence of conformational dynamics on enzymatic networks and then presents a broad perspective on statistical kinetics and nonequilibrium thermodynamics of driven processes, including enzyme turnover, molecular motor translocation, ion transport, and fluorescence emission. Single enzyme experiments reveal conformational fluctuations on multiple time scales, but the measured turnover rate still follows the simple Michaelis-Menten (MM) equation. To resolve this puzzle, we establish the generalized rate equation for an arbitrary enzymatic network, which reduces to the MM form when conformational detailed balance is obeyed. The generalized MM expression reveals a relation between non-MM corrections and non-equilibrium conformational currents and predicts a resonant effect in the speed-up of enzyme turnovers. If time allows, I will also discuss the detection of such non-equilibrium currents as well as the information content of the Poisson indicator, which quantifies the statistics of chemical reactions. Our analysis of the enzymatic networks can be extended to a general study of kinetics and nonequilibrium thermodynamics of driven biomolecular systems. (i) Technically, a pathway approach allows us to decompose and combine complex chemical networks using kinetic motifs. (ii) Conceptually, for a dynamically disordered biomolecular machine regulated by a hidden process, a time-based fluctuation theorem no longer applies to the observable first-passage time; however, we find that its validity is restored in the absence of hidden currents (e.g., conformational detailed balance). (iii) Our analysis of chemical networks and the fluctuation theorem shed new light on the validity of the course-graining procedure widely adopted in molecular modelling.