Dr. James (Junzhuo) Liao
Tel: (516) 461-9615
![Junzhuo_Liao-e1624191011617-300x300-c.jpeg[1].webp](https://static.wixstatic.com/media/c5eede_7ae39ff0fd494c32a88421b9acc8ce0a~mv2.png/v1/fill/w_113,h_134,al_c,q_85,usm_0.66_1.00_0.01,enc_avif,quality_auto/Junzhuo_Liao-e1624191011617-300x300-c_jpeg%5B1%5D.png)
Columbia University
New York, NY
Postdoctoral Research, Computer Simulations for CADD (Computer-Aided Drug Design)
(Sept. 2022 - Current )
University of Wisconsin-Madison Madison, WI
Postdoctoral Research, Computational Medicinal Chemistry
(Dec. 2019 - Aug. 2022)
Stony Brook University
Stony Brook, NY
PhD, Chemistry (2017)
Peking University
Beijing, China
BS, Pharmaceutical Sciences (2010)
Education and Academic Work
Published Peer-Reviewed Journal Articles
First Author or Co-first Author
-
Junzhuo Liao, Alina P. Sergeeva, Edward D. Harder, Lingle Wang, Jared M. Sampson, Barry Honig, Richard A. Friesner, “A Method for Treating Significant Conformational Changes in Alchemical Free Energy Simulations of Protein–Ligand Binding.” J. Chem. Theory Comput. 2024, 20, 19, 8609–8623. https://doi.org/10.1021/acs.jctc.4c00954 (Cover article)
-
Junzhuo Liao, Xueqing Nie, Ilona Christy Unarta, Spencer S. Ericksen and Weiping Tang, “Modeling and Scoring of PROTAC-Mediated Ternary Complex Poses.” J. Med. Chem. 2022, 65, 8, 6116–6132. https://doi.org/10.1021/acs.jmedchem.1c02155 Online April 12, 2022.
-
Daniel A. Glazier,† Junzhuo Liao,† Brett L. Roberts,† Xiaolei Li, Ka Yang, Christopher M. Stevens and Weiping Tang, “Chemical Synthesis and Biological Application of Modified Oligonucleotides.” Bioconjugate Chem. 2020, 31, 5, 1213–1233. https://doi.org/10.1021/acs.bioconjchem.0c00060 †co-first author
-
Junzhuo Liao and Dale G. Drueckhammer, “Diastereoselective conjugate addition/cyclization/bromination: Access to four stereocenters in a single step.” Tetrahedron Lett., 2018, 59, 1776-1778. https://doi.org/10.1016/j.tetlet.2018.03.080
Other Co-authorships
-
The Atomwise AIMS Program (Junzhuo Liao, et. al.), “AI is a viable alternative to high throughput screening: a 318-target study.” Sci. Rep., 2024, 14, 7526 (2024) https://doi.org/10.1038/s41598-024-54655-z
-
Xueqing Nie†, Yu Zhao†, Hua Tang†, Zhongrui Zhang, Junzhuo Liao, Chelsi M. Almodovar-Rivera, Ramya Sundaresan, Haibo Xie, Le Guo, Bo Wang, Hongqing Guan, Yongna Xing, Weiping Tang, “Development of Phenyl-substituted Isoindolinone- and Benzimidazole-type Cereblon Ligands for Targeted Protein Degradation.” Chembiochem, 2024, 25, e2023006. https://doi.org/10.1002/cbic.202300685 †co-first author
-
Alina P. Sergeeva, Phinikoula S. Katsamba, Junzhuo Liao, Jared M. Sampson, Fabiana Bahna, Seetha Mannepalli, Nicholas C. Morano, Lawrence Shapiro, Richard A. Friesner and Barry Honig, “Free Energy Perturbation Calculations of Mutation Effects on SARS-CoV-2 RBD::ACE2 Binding Affinity.” J. Mol. Biol. 2023, 435 (15), 168187. https://doi.org/10.1016/j.jmb.2023.168187
-
Jingyao Li†, Chunrong Li†, Zhongrui Zhang†, Zhen Zhang†, Zhiping Wu, Junzhuo Liao, Zhen Wang, Meghan McReynolds, Haibo Xie, Le Guo, Qiuhua Fan, Junmin Peng, Weiping Tang, “A platform for the rapid synthesis of molecular glues (Rapid-Glue) under miniaturized conditions for direct biological screening.” Eur. J. Med. Chem. 2023, 258, 115567. https://doi.org/10.1016/j.ejmech.2023.115567 †co-first author
-
Haibo Xie†, Chunrong Li†, Tang Hua, Ira Tandon, Junzhuo Liao, Brett Roberts, Yu Zhao, Weiping Tang, “Development of Substituted Phenyl Dihydrouracil as the Novel Achiral Cereblon Ligands for Targeted Protein Degradation.” J. Med. Chem. 2023, 66, 4, 2904-2917. https://doi.org/10.1021/acs.jmedchem.2c01941 †co-first author
-
Dale G. Drueckhammer, Steven Qizhi Gao, Xiaofei Liang and Junzhuo Liao, “Acetone–Heptane as a Solvent System for Combining Chromatography on Silica Gel with Solvent Recycling.” ACS Sustainable Chem. Eng., 2013, 1, 87–90. https://doi.org/10.1021/sc300044c (Founding issue)
Other Publications and Presentations
-
Nataly Cruz-Rodriguez, Hua Tang, Milad Rouhimoghadam, Junzhuo Liao, Benjamin Bateman, Briana Bates, Mario Pinzas, Xuhui Huang, Weiping Tang, Michael W. Deininger; “Targeting Chronic Myeloid Leukemia with Potent and Specific BCR::ABL1 Degraders” Blood 2024, 144, 4157. https://doi.org/10.1182/blood-2024-211207
-
Milad Rouhimoghadam, Hua Tang, Briana Bates, Junzhuo Liao, Daniela Uribe-Cano, Weiping Tang, Michael W. Deininger “Discovery of a Selective PROTAC Degrading Untamable BCR::ABL1 Compound Mutants in Chronic Myelogenous Leukemia (CML)and Philadelphia Chromosome-Positive (Ph+) Acute Lymphoblastic Leukemia (ALL).” Clin. Lymphoma Myeloma Leuk. 2023, 23, S336. https://doi.org/10.1016/S2152-2650(23)01131-X
-
Milad Rouhimoghadam, Hua Tang, Junzhuo Liao, Briana Bates, Daniela Uribe-Cano, Helong Zhao, Weiping Tang, Michael W. Deininger; “LPA81: Discovery of an Exceptionally Potent Protac Degrading Native and Mutant BCR-ABL1 Oncoprotein in CML.” Blood 2022, 140 (Supplement 1): 485–486. https://doi.org/10.1182/blood-2022-158308
-
Junzhuo Liao, Haibo Xie, Spencer S. Ericksen and Weiping Tang, “Structure-Based Virtual Screening for an E3 Ligase Binder from Enamine Libraries.” Poster Presentation, 58th Annual MIKIW Meeting, Madison, WI, April 24-25, 2021
-
Junzhuo Liao, Xiao Liu and Dale Drueckhammer, “Design and Synthesis of a Novel Nucleic Acid Mimic.” Poster Presentation, 248th National Meeting of the American Chemical Society, San Francisco, CA, August 10–14, 2014.
-
Junzhuo Liao, “Design and Synthesis of a Novel Nucleic Acid Mimic.” Doctoral dissertation, 2017, The Graduate School, Stony Brook University: Stony Brook, NY.
Inventions
-
Junzhuo Liao, Robert William Weir, Xiao Yu, Zhanjun Luo, “Live-Column Visualization Chromatography for Separation of Compounds” (Non-provisional Patent, filed 5/16/2023, pending)
-
Weiping Tang, Haibo Xie, Yunxiang Wu, Junzhuo Liao, Chunrong Li, Ira Tandon, and Hua Tang, “Synthesis of Novel Cereblon E3 Ligase Ligands, Compounds Formed Thereby, and Pharmaceutical Compositions Containing Them” (Non-provisional Patent, filed 1/30/2024, pending)
Reviewer Activity
-
Served as reviewer on 2 manuscripts for Crit. Rev. Biotechnol., 2022.
Research Experience
-
Postdoc projects:
-
UW - Madison (Dec. 2019 – Aug. 2022)
-
-Modeling and Scoring of PROTAC-Mediated Ternary Complex Poses. Scripts and data available on github.com/JL2021MD/PROTACModeling.
See also https://pubs.acs.org/doi/abs/10.1021/acs.jmedchem.1c02155 and a summary at the end of this CV.
-Modeling of BCR/ABL-PROTAC-CRBN ternary complexes and its ubiquitination process (adapting above success of general PROTAC modeling).
-Explicit solvent binding enthalpy calculations. Adapting the idea from the MM/GBSA method, can we easily calculate enthalpies of a static solute, and obtain binding enthalpies in explicit solvent? (This project evolved into single molecule pulling proposal)
-Scoring of ligand-receptor binding affinity in explicit solvent by heating for virtual screening. Stronger binding ligands will be harder to dissociate, and vice versa. A correct binding pose as the starting point is very important, however. (This project evolved into single molecule pulling proposal)
-Modeling of various protein-ligand interactions (STAT3, FGFR, DCAF, CRBN-IKZF1, CRBN-GSPT1, CD206, VHL)
-Large scale virtual screening of ligands to a cryptic binding site on CRBN
(Non-modeling projects below)
-High-throughput screening of CHIP binding small molecules, as well as biophysical binding measurements of various ligands to their target proteins, by fluorescence polarization (FP), microscale thermophoresis (MST), and differential scanning fluorimetry (DSF).
-Oligonucleotide synthesis for OligoTac, a selective degrader of Ago2 based on its associated RNA sequence. With Ago2 degraded by the ubiquitin pathway, the exposed miRNA will therefore be naturally degraded.
Advisor: Dr. Weiping Tang
-
PhD project on the de novo design, virtual screening, synthesis and evaluation of novel nucleic acid mimics-molecules that chiefly serve antisense functions. Numerous computationally designed (with HostDesigner) nucleic acid mimics were subjected to molecular dynamics simulations and estimations of binding affinity with natural DNA/RNA. The mimic that had the best binding results was prioritized for synthesis. The synthetic route on paper was very challenging, but the structure was also most innovative, having a backbone where large phenanthrene rings were inserted in between natural nucleobases resulting in an expanded helix structure, and the novelty was also an important driving force. Eventually after years of long synthetic chemistry work, a short oligomer was successfully made. Limited binding affinity tests with the tiny amount of material to a complementary RNA strand were done via isothermal titration calorimetry (ITC). After my leaving, due to lack of future personnel and funding, this project was briefly continued in my postdoc group until other projects took over.
Advisor: Dr. Dale G. Drueckhammer
A research summary that specifically focuses on the computational modeling part of this project is provided at the end of this CV, while part of the chemistry is reported in Tetrahedron. Lett.
-
Undergraduate dissertation project on the synthetic medicinal chemistry studies of novel dithiocarbamate EGFR tyrosine kinase inhibitors.
Advisor: Dr. Runtao Li
Current Preference of Computational Modeling Engines and Coding
AMBER MD and AmberTools (11, 18-20), UCSF Chimera, Docking (Smina, Surflex, GOLD, rDock), Linux Bash and Python.
Teaching Experience
All the following assignments are teaching assistant positions in Stony Brook University. I also enjoy teaching and training others in lab, because it directly helps others while I improve in the process too.
-
CHE 322: Organic Chemistry II, Spring 2017.
-
CHE 327: Organic Chemistry Lab, Fall 2010, 2012, 2016, Spring 2011, 2013, 2015-16.
-
CHE 384: Intermediate Synthetic and Spectroscopic Lab Techniques, Fall 2014-15, Spring 2014.
-
CHE 133: General Chemistry Lab I, Summer 2014-16.
-
CHE 131: General Chemistry I, Summer 2014.
-
CHE 132: General Chemistry II, Fall 2013, Summer 2014.
-
Responsible of new user training for all campus-wide users of the MST instrument (2021-22, UW-Madison.)
Instrument Experience
-
Microscale Thermophoresis (MST), Fluorescence Polarization (FP), Differential Scanning Fluorimetry (DSF), SPR and OpenSPR. Responsible of new user training for all campus-wide users of the MST.
-
Mass spectrometry (Agilent 1100 LC-MS used frequently). Written original research proposal “A Convenient Preparative Mass Spectrometer with a Liquid Phase Soft Landing Surface for Separation of Organic Compounds” as a mimic for grant writing, part of my third committee examination towards the PhD degree.
-
Fully proficient with the automated column chromatograph (Combiflash Rf), and had experiences of opening up the device for simple repairs. Varian 300MHz NMR spectrometer, Bruker 300MHz, 400MHz and 700MHz NMR spectrometers used for daily research. FT-IR spectrometer, UV-Vis spectrometer, prep TLC and HPLC also used.
Research Summaries
Modeling work on predicting native PROTAC-ternary complexes
PROTAC is a small molecule which one end binds to an E3 ligase, and the other end binds to a target protein, causing the target protein to be degraded via the ubiquitin pathway. The E3 ligase leads towards recruitment of ubiquitin. Knowing the correct ternary pose of the protein-PROTAC-protein system is a fundamental step in performing structure-based PROTAC design, and also atomic-level 3D understanding of the ubiquitination process.
Using MD to recapitulate the solved cocrystal structures was successful, if using the exact binary structures as starting point, even though the search and pose selection process was not very efficient. The prediction work can be separated into 2 stages: 1.coming up with candidate poses, 2.distinguishing the correct pose. Ideally, 2 proteins and PROTAC should directly form the native pose as the dominant pose, so there are not 2 separate stages.
However under many occasions the most populous pose in simulation isn’t the native one found from crystal. One explanation is that the crystal pose isn’t the lowest free energy pose in solution, the other common explanation is insufficient sampling, though I feel those may not be a sufficient answer. For example this recent PROTAC simulation paper, the native state based on their very extensive MD sampling is only ranked the 4th lowest in free energy (Figure 8a).
Figure 8a in their paper. The red X is the crystal native state. Blue dots are found states with experimental info guidance towards the native crystal included. There’s not much explanation on the found lower energy states, nor evidence of solution structures corresponding to them.
Without evidence of corresponding solution poses, I think this could be related to the force field, and an example can be seen from a 1995 paper published by Berne and Friesner that 2 FFs give the same conformations very different energy ranks, Tables 5-7 in paper. Another example can be seen from a 2007 paper by Donald and Shakhnovich et. al., that Figures 2,4 in paper shows how different potentials can give different spatial features and energies.
Here’s my analysis from a hypothetical example: small inaccuracies in bond and dihedral angle parameters over a long time and sufficient number of runs could cause the dominant pose to be different from the true native (a propagation of error). But when native binding pose determination is not based on population, but based on FEP, TI, or even MMGBSA or LIE, intramolecular energy of each candidate pose is canceled out or not considered, and such minor bond inaccuracies won’t affect the results.
Those are my understandings so far on why a single stage prediction of PROTAC ternary poses are very difficult, and therefore in my study there are 2 separate stages.
Table of Contents Summary Figure in Our Paper.
For the next stage, MMGBSA was used at first. It was indeed useful and the native pose was favored, but not all the time. This could be due to one or both of the following reasons 1.inaccuracy in the implicit solvent MMGBSA model and 2.not considering entropy. Eventually, I came up with a simple-to-use scoring method that avoids a separate set of GB parameters, in explicit water, and that naturally considers any entropic effects (Summarized in TOC Figure above). It worked very well to distinguish native vs non-native poses; and it also provides empirical guidance on whether a given pose is likely native or not, when no other method published so far can provide any kind of hint on this important question. When scoring a series of candidate poses and the native one isn’t even included in it, if the scoring method can tell the researcher this information, he/she will know to try search for more candidate poses. Compare this to Rosetta and its magnificent Rosetta@Home, imagine only a small number of users contributed some protein structures to the server, there will always be a best scoring one, but we don’t know if any of the small number of protein structures contain the native one. But if we know, we can continue or stop the search.
While our work was being conducted, several papers on this topic appeared that reported ternary docking methods. They were able to search the conformation space adequately, and the native pose was almost always included in the results. But their distinguishing part didn’t work well. One of them, PRosettaC was readily available as an online server, and we also scored their poses. The crystal native pose was always the best, with negative ones all scoring worse than it, the ideal situation. I made PRosettaC and our scoring method more compatible and practical to a user by adding a quick refinement step to gently move PRosettaC poses more towards the bottom of the energy well in the FF our scoring was using. This work was first published in Nov. last year in Chemrxiv, with several updates afterwards, and just April 2022 the newest version had been successfully published in J. Med. Chem.
In addition to conjugation with PRosettaC, we’re now able to use it in conjugation with the more recent Integrative PROTAC-Model for the stage of quickly generating ternary docking poses.
Computational modeling work on novel nucleic acids as part of PhD project
Numerous types of de novo design nucleic acid mimics underwent molecular dynamics modeling with AMBER, for simulation of binding with natural DNA/RNA and estimation of binding energy. These mimics which have fully new backbones but natural nucleobases were designed based on spatial fitting with HostDesigner. The design goal was the mimic having stronger binding affinity to its natural complementary strand and having nuclease resistance, compared to natural DNA/RNA duplexes. The general AMBER force field (GAFF) was used to parameterize new molecules (designed mimics), and either a single parameterization or residue-wise parameterization method was used (I’ve written out the parameterization method on the Rizzo Lab Wiki, if scrolled to nucleic acids section. As of today, this is a bit outdated and I’ve worked out a better method: using AmberTools program prepgen, different monomeric units are able to be converted into prep files, just like how amino acid and nucleotide residues are stored in the AMBER library. Before running prepgen, one should first assign some high-quality charges like AM1-BCC charges to each capped monomeric unit, like a small molecule ligand. GAFF was used to generate all the bond parameters.). Each 9 or 12-unit long nucleic acid mimic strand with its natural complementary underwent a 2 ns length, explicit-solvent binding simulation followed by MMGBSA binding energy estimations.
The mimic with the best binding score was one that had by design an expanded helix structure where large phenanthrene rings were inserted in between natural nucleobases (Figure 1 right and Figure 2 bottom center), for extra π-stacking binding force. This mimic was chosen for synthesis.
Figure 1. Various designs of nucleic acid mimics. The left series is based on changing the pentose ring into a furan ring for better rigidity, with suitable linkers connecting each unit. The right series is based on an expanded helix structure, with large aromatic rings inserted in between from the complementary strand. A snapshot from the binding simulation trajectory of the chosen target for synthesis is shown.
Figure 2. Base pair distance stability analysis for several designs and natural duplex controls. Binding enthalpy by MMGBSA shown at top right. The SBNAnOMe design, which was also most innovative structure-wise, had the best stability and was the prioritized synthesis target.