“Zhanglong is a passionate scientist who devotes to make a difference in his research area. He is hard-working, diligent and self-motivated. I strongly recommend him and believe he will be a great scientist and leader in his future career. ”
John (Zhanglong) Liu
Fremont, California, United States
630 followers
500+ connections
Contributions
-
What do you do if your machine learning models lack logical reasoning abilities?
It is not a surprise at all if a bad model lacks reasoning ability. With a good model training and evaluation, if a model still lacks reasoning ability, it is good to check the feature leakage, as it will make training and testing metrics perfect. As the model objective or future data is leaked in the feature, the model will not learn the expected reasoning logic well.
-
What are the most important things to do when starting a new job in Machine Learning?
Try to study the innovations and problems of projects. The innovation may help spark new ideas in the future models/projects, as lots of knowledge and learning are transferable. While, the problems will help initiate new directions.
-
What are some methods for exploring and analyzing your data before applying machine learning?
Data analysis is one of the most important steps of the Machine Learning modeling. 1) once the project goal was set up, we could use analysis to check whether the model objectives are aligned with the project goals. 2) once the model objective is fixed, the analysis could help feature selection. 3) analysis could also help simulate the model impacts in the future.
Activity
-
Had a fun chat with the team at Andreessen Horowitz, sharing a behind-the-scenes look at our foray into the games space — and how it all ties back to…
Had a fun chat with the team at Andreessen Horowitz, sharing a behind-the-scenes look at our foray into the games space — and how it all ties back to…
Liked by John (Zhanglong) Liu
-
If you missed the previous version of our webinar, and would like to know about how Kumo.AI can help you deliver better, cheaper, faster results on…
If you missed the previous version of our webinar, and would like to know about how Kumo.AI can help you deliver better, cheaper, faster results on…
Liked by John (Zhanglong) Liu
-
Similar to the trend we are seeing in our data as well… Shout out to a few of the many CEOs doing it right on LinkedIn: Chris Kempczinski Patrice…
Similar to the trend we are seeing in our data as well… Shout out to a few of the many CEOs doing it right on LinkedIn: Chris Kempczinski Patrice…
Liked by John (Zhanglong) Liu
Experience
Education
-
Stanford University
-
Activities and Societies: Linear Regression on House Price problem, Logistic Regression on Cancer prediction, Neural Network on Character Recognition, Image Compression by PCA, K-means Clustering, Support Vector Machine, Anomaly Detection, Recommender System, Pipeline Design, Large Scale Machine Learning (Batch, Mini-batch and Stochastic gradient descent)
-
-
Licenses & Certifications
Volunteer Experience
-
Teaching Assistant
University of Florida
- 1 year 7 months
Science and Technology
Instructed undergraduate students working on Molecular Conformation Equilibrium by NMR, Data processing and interpretation, Heat capacity, Colligative properties, Phase diagram, Fluorescence/ Quenching/Energy transfer;
Publications
-
Effects of Hinge Region Natural Polymorphisms on Human Immunodeficiency Virus-1 Protease Structure, Dynamics and Drug-Pressure Evolution
J. Biol. Chem.
Multidrug resistance to current FDA approved HIV-1 protease (PR) inhibitors drives the need to understand the fundamental mechanisms of how drug-pressure selected mutations, which are oftentimes natural polymorphisms, elicit their effect on enzyme function and resistance. Here, the impact of hinge-region natural polymorphism at residue 35 - glutamate to aspartate (E35D) - alone and in conjunction with residue 57 arginine to lysine (R57K) are characterized with the goal of understanding how…
Multidrug resistance to current FDA approved HIV-1 protease (PR) inhibitors drives the need to understand the fundamental mechanisms of how drug-pressure selected mutations, which are oftentimes natural polymorphisms, elicit their effect on enzyme function and resistance. Here, the impact of hinge-region natural polymorphism at residue 35 - glutamate to aspartate (E35D) - alone and in conjunction with residue 57 arginine to lysine (R57K) are characterized with the goal of understanding how altered salt-bridge interactions between the hinge and flap regions are associated with changes in structure, motional dynamics, conformational sampling, kinetic parameters, and inhibitor affinity. The combined results reveal that the single E35D substitution leads to diminished salt-bridge interactions between residue 35 and 57 and gives rise to the stabilization of open-like conformational states with overall increased backbone dynamics. In HIV-1 PR constructs where sites 35 and 57 are both mutated (e.g. E35D, R57K), X-ray structures reveal an altered network of interactions that replace the salt-bridge thus stabilizing the structural integrity between the flap and hinge regions. In spite of the altered conformational sampling and dynamics when the salt bridge is disrupted, enzyme kinetic parameters and inhibition constants are similar to those obtained for subtype B PR. Results demonstrate that these hinge region natural polymorphisms, which may arise as drug pressure secondary mutations, alter protein dynamics and conformational landscape, which are important thermodynamic parameters to consider for development of inhibitors that target for non-subtype B PR.
Other authorsSee publication -
Pulsed EPR Characterization of HIV-1 Protease Conformational Sampling and Inhibitor-Induced Population Shifts
PHYSICAL CHEMISTRY CHEMICAL PHYSICS
The conformational landscape of HIV-1 protease (PR) can be experimentally characterized by pulsed-EPR double electron-electron resonance (DEER). For this characterization, nitroxide spin labels are attached to an engineered cysteine residue in the flap region of HIV-1 PR. DEER distance measurements from spin-labels contained within each flap of the homodimer provide a detailed description of the conformational sampling of apo-enzyme as well as induced conformational shifts as a function of…
The conformational landscape of HIV-1 protease (PR) can be experimentally characterized by pulsed-EPR double electron-electron resonance (DEER). For this characterization, nitroxide spin labels are attached to an engineered cysteine residue in the flap region of HIV-1 PR. DEER distance measurements from spin-labels contained within each flap of the homodimer provide a detailed description of the conformational sampling of apo-enzyme as well as induced conformational shifts as a function of inhibitor binding. The distance distribution profiles are further interpreted in terms of a conformational ensemble scheme that consists of four unique states termed "curled/tucked", "closed", "semi-open" and "wide-open" conformations. Reported here are the DEER results for a drug-resistant variant clinical isolate sequence, V6, in the presence of FDA approved protease inhibitors (PIs) as well as a non-hydrolyzable substrate mimic, CaP2. Results are interpreted in the context of the current understanding of the relationship between conformational sampling, drug resistance, and kinetic efficiency of HIV-1PR as derived from previous DEER and kinetic data for a series of HIV-1PR constructs that contain drug-pressure selected mutations or natural polymorphisms. Specifically, these collective results support the notion that inhibitor-induced closure of the flaps correlates with inhibitor efficiency and drug resistance. This body of work also suggests DEER as a tool for studying conformational sampling in flexible enzymes as it relates to function.
Other authorsSee publication -
Electron Spin Resonance Scanning Probe Spectroscopy for Ultra-Sensitive Biochemical Studies
Analytical Chemisty
Electron spin resonance (ESR) spectroscopy's affinity for detecting paramagnetic free radicals, or spins, has been increasingly employed to examine a large variety of biochemical interactions. Such paramagnetic species are broadly found in nature and can be intrinsic (defects in solid-state materials systems, electron/hole pairs, stable radicals in proteins) or, more often, purposefully introduced into the material of interest (doping/attachment of paramagnetic spin labels to biomolecules of…
Electron spin resonance (ESR) spectroscopy's affinity for detecting paramagnetic free radicals, or spins, has been increasingly employed to examine a large variety of biochemical interactions. Such paramagnetic species are broadly found in nature and can be intrinsic (defects in solid-state materials systems, electron/hole pairs, stable radicals in proteins) or, more often, purposefully introduced into the material of interest (doping/attachment of paramagnetic spin labels to biomolecules of interest). Using ESR to trace the reactionary path of paramagnetic spins or spin-active proxy molecules provides detailed information about the reaction's transient species and the label's local environment. For many biochemical systems, like those involving membrane proteins, synthesizing the necessary quantity of spin-labeled biomolecules (typically 50 pmol to 100 pmol) is quite challenging and often limits the possible biochemical reactions available for investigation. Quite simply, ESR is too insensitive. Here, we demonstrate an innovative approach that greatly enhances ESR's sensitivity (> 20 000 times improvement) by developing a near-field, non-resonant, X-band ESR spectrometric method. Sensitivity improvement is confirmed via measurement of 140 amol of the most common nitroxide spin label in a ≈ 593 fL liquid cell at ambient temperature and pressure. This experimental approach eliminates many of the typical ESR sample restrictions imposed by conventional resonator-based ESR detection and renders the technique feasible for spatially resolved measurements on a wider variety of biochemical samples. Thus our approach broadens the pool of possible biochemical and structural biology studies as well as greatly enhances the analytical power of existing ESR applications.
Other authors -
Continuous wave W- and D-Band EPR spectroscopy offer “sweet-spots” for characterizing conformational changes and dynamics in intrinsically disordered proteins
Biochemical and Biophysical Research Communications
Site-directed spin labeling (SDSL) electron paramagnetic resonance (EPR) spectroscopy is a powerful tool for characterizing conformational sampling and dynamics in biological macromolecules. Here we demonstrate that nitroxide spectra collected at frequencies higher than X-band (∼9.5 GHz) have sensitivity to the timescale of motion sampled by highly dynamic intrinsically disordered proteins (IDPs). The 68 amino acid protein IA3, was spin-labeled at two distinct sites and a comparison of X-band…
Site-directed spin labeling (SDSL) electron paramagnetic resonance (EPR) spectroscopy is a powerful tool for characterizing conformational sampling and dynamics in biological macromolecules. Here we demonstrate that nitroxide spectra collected at frequencies higher than X-band (∼9.5 GHz) have sensitivity to the timescale of motion sampled by highly dynamic intrinsically disordered proteins (IDPs). The 68 amino acid protein IA3, was spin-labeled at two distinct sites and a comparison of X-band, Q-band (35 GHz) and W-band (95 GHz) spectra are shown for this protein as it undergoes the helical transition chemically induced by tri-fluoroethanol. Experimental spectra at W-band showed pronounced line shape dispersion corresponding to a change in correlation time from ∼0.3 ns (unstructured) to ∼0.6 ns (α-helical) as indicated by comparison with simulations. Experimental and simulated spectra at X- and Q-bands showed minimal dispersion over this range, illustrating the utility of SDSL EPR at higher frequencies for characterizing structural transitions and dynamics in IDPs.
Other authors -
Polarity-manipulation based on nanoscale structural transformation on strained 2D MgO
Journal of Physics D: Applied Physics
Strain induced nanoscale structural transformation is demonstrated in this paper to have the ability of triggering polarity flipping in a wide bandgap system of MgxZn1−xO/MgO/Al2O3. Relaxation dynamics of semiconductor components under large compressive pressures up to 13.7 GPa were studied by a combination of theoretical analysis and experimental characterizations including in situ reflection high-energy electron diffraction and high-resolution transmission electron microscopy. The gigantic…
Strain induced nanoscale structural transformation is demonstrated in this paper to have the ability of triggering polarity flipping in a wide bandgap system of MgxZn1−xO/MgO/Al2O3. Relaxation dynamics of semiconductor components under large compressive pressures up to 13.7 GPa were studied by a combination of theoretical analysis and experimental characterizations including in situ reflection high-energy electron diffraction and high-resolution transmission electron microscopy. The gigantic force between MgZnO and ultrathin-MgO/Al2O3 delayed the structural transformation of MgZnO from six-fold cubic to four-fold wurtzite into the second monolayer, and consequently flipped the polarity of the film deposited on relaxed MgO. Additionally, dislocation-induced strain relaxation was suggested to happen around 1 nm thick cubic MgO grown on Al2O3, instead of the previous well-accepted concept that wurtzite structures can be inherited from the oxygen sub-lattice of sapphire substrates below the critical thickness. Finally, the structural transformation method employing an ultrathin-MgO interfacial layer was demonstrated to be a suitable technique for accommodating the large lattice mismatch comparing with the dislocation-relaxation mechanism achieving a UV photodetector with four orders of rejection ratio of the UV-to-visible photoresponse.
Other authorsSee publication -
Towards Increased Concentration Sensitivity for Continuous Wave EPR Investigations of Spin-Labeled Biological Macromolecules at High Fields
journal of magnetic resonance
High-field, high-frequency electron paramagnetic resonance (EPR) spectroscopy at W- (~95 GHz) and D-band (~140 GHz) is important for investigating the conformational dynamics of flexible biological macromolecules because this frequency range has increased spectral sensitivity to nitroxide motion over the 100 ps to 2 ns regime. However, low concentration sensitivity remains a roadblock for studying aqueous samples at high magnetic fields. Here, we examine the sensitivity of a non-resonant…
High-field, high-frequency electron paramagnetic resonance (EPR) spectroscopy at W- (~95 GHz) and D-band (~140 GHz) is important for investigating the conformational dynamics of flexible biological macromolecules because this frequency range has increased spectral sensitivity to nitroxide motion over the 100 ps to 2 ns regime. However, low concentration sensitivity remains a roadblock for studying aqueous samples at high magnetic fields. Here, we examine the sensitivity of a non-resonant thin-layer cylindrical sample holder, coupled to a quasi-optical induction-mode W-band EPR spectrometer (HiPER), for continuous wave (CW) EPR analyses of: (i) the aqueous nitroxide standard, TEMPO; (ii) the unstructured to -helical transition of a model IDP protein; and (iii) the base-stacking transition in a kink-turn motif of a large 232 nt RNA. For sample volumes of ~50 L, concentration sensitivities of 2-20 µM were achieved, representing a ~10-fold enhancement compared to a cylindrical TE011 resonator on a commercial Bruker W-band spectrometer. These results therefore highlight the sensitivity of the thin-layer sample holders employed in HiPER for spin-labeling studies of biological macromolecules at high fields, where applications can extend to other systems that are facilitated by the modest sample volumes and ease of sample loading and geometry.
Other authors
Patents
-
Deep Representation Machine Learned Model for Heterogeneous Information Networks
11,941,057
Courses
-
Introduction to Data Science in Python by University of Michigan
coursera
-
Machine Learning by Stanford University
coursera
-
Python for Data Science
UCSanDiegoX: DSE200x
-
R Programming by Johns Hopkins University
Coursera
Honors & Awards
-
Outstanding Research Award, National Synchrotron Radiation Laboratory, China
-
-
Outstanding Award (Top 10%), Institute of Physics, UCAS
-
-
Recognition Award (Top 30%), Institute of Physics, UCAS
-
-
Graduate Research Fellowship from Institute of Physics
-
This is a five-year Graduate Research Fellowship from 2006 to 2011
Languages
-
English
Full professional proficiency
-
Chinese
Native or bilingual proficiency
Organizations
-
Biophysical Society
-
- Present
Recommendations received
1 person has recommended John (Zhanglong)
Join now to viewMore activity by John (Zhanglong)
-
It's been a couple years since I graduated from college, but I still remember how busy back-to-school time was — new classes, new friends, and…
It's been a couple years since I graduated from college, but I still remember how busy back-to-school time was — new classes, new friends, and…
Liked by John (Zhanglong) Liu
-
A lesson worth remembering - people work with managers and leaders who not only provide them with the best opportunities and growth, but also help…
A lesson worth remembering - people work with managers and leaders who not only provide them with the best opportunities and growth, but also help…
Liked by John (Zhanglong) Liu
-
Cheers to 8 years! 🎉 Grateful for seeing how much I’ve grown in my professional career by achieving personal milestones, making mistakes and…
Cheers to 8 years! 🎉 Grateful for seeing how much I’ve grown in my professional career by achieving personal milestones, making mistakes and…
Liked by John (Zhanglong) Liu
-
Hi friends and colleagues, After an amazing and transformative 10 years at LinkedIn, I’m ready to try something new. I’m looking to apply my data…
Hi friends and colleagues, After an amazing and transformative 10 years at LinkedIn, I’m ready to try something new. I’m looking to apply my data…
Liked by John (Zhanglong) Liu
-
In our paper is presented at KDD 2024: LiRank: Industrial Large Scale Ranking Models at LinkedIn (https://lnkd.in/g7djGN9y), we discuss the…
In our paper is presented at KDD 2024: LiRank: Industrial Large Scale Ranking Models at LinkedIn (https://lnkd.in/g7djGN9y), we discuss the…
Liked by John (Zhanglong) Liu
Other similar profiles
Explore collaborative articles
We’re unlocking community knowledge in a new way. Experts add insights directly into each article, started with the help of AI.
Explore More