Qing Shao
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Qing Shao has a total of 172 co-authors. Here are the top co-authors by number of shared publications:
- Qi Qiao
Department of Biostatistics, University of Kentucky College of Public Health, Lexington, USA
16 shared publications - Xiaoqi Liu
Department of Toxicology and Cancer Biology, University of Kentucky
14 shared publications - Anju Karki
Cardiff Oncology Inc, San Diego, CA
13 shared publications - Bernard Mark Evers
University of Kentucky
13 shared publications - Chi Wang
Markey Cancer Center, University of Kentucky, Lexington, KY 40536, USA
13 shared publications - Daheng He
Markey Cancer Center, University of Kentucky
13 shared publications
Latest Publications
- Open Access
Benchmark Averages Hide the Failures That Matter: Quantizing ESM-2 for Protein Variant-Effect Prediction
We benchmark six numerical precision configurations for ESM-2 protein language models across throughput, memory footprint and predictive accuracy, on two workloads with sharply different characteristics: bulk embedding extraction and deep mutational scanning (DMS) variant-effect scoring. Accuracy is evaluated on the complete ProteinGym substitution benchmark — 201 assays, 2.41M variants — at three model scales spanning 650M to 15B parameters , with a paired bootstrap clustered on protein. Three findings follow, and each contradicts a common practice. First, benchmark averages conceal the failure that decides deployability : no configuration shifts mean correlation by more than 0.007 at any scale, yet INT8 dynamic quantization — indistinguishable from fp32 on that mean at 3B ( p = 0.34) — takes a single assay from ρ = 0.591 to 0.223. Selection must be made on worst-case, not mean, behaviour. Second, fidelity measured against fp32 bounds risk but cannot rank quality : over 3015 assay/configuration pairs it predicts the magnitude of ground-truth change ( r = 0.56–0.81) but not its direction , and the INT4 effect differs significantly between 650M and 3B (+0.0101, p = 0.0007) with no monotone trend to extrapolate. Third, quantizing a large model is dominated by using a small one : of eighteen scale/configuration combinations only three are Pareto-optimal over accuracy, memory and speed, and all three are 650M. The one catastrophic failure we observe is a defect of default symmetric activation scaling, not of W8A8 itself: asymmetric activation quantization, a one-line change needing no calibration, removes every damaged assay. We also give a label-free screen for at-risk targets, and report four measurement artifacts encountered during this study, three of which inverted the result they were meant to measure.
Open AccessBenchmark Averages Hide the Failures That Matter: Quantizing ESM-2 for Protein Variant-Effect Prediction
We benchmark six numerical precision configurations for ESM-2 protein language models across throughput, memory footprint and predictive accuracy, on two workloads with sharply different characteristics: bulk embedding extraction and deep mutational scanning (DMS) variant-effect scoring. Accuracy is evaluated on the complete ProteinGym substitution benchmark — 201 assays, 2.41M variants — at three model scales spanning 650M to 15B parameters , with a paired bootstrap clustered on protein. Three findings follow, and each contradicts a common practice. First, benchmark averages conceal the failure that decides deployability : no configuration shifts mean correlation by more than 0.007 at any scale, yet INT8 dynamic quantization — indistinguishable from fp32 on that mean at 3B ( p = 0.34) — takes a single assay from ρ = 0.591 to 0.223. Selection must be made on worst-case, not mean, behaviour. Second, fidelity measured against fp32 bounds risk but cannot rank quality : over 3015 assay/configuration pairs it predicts the magnitude of ground-truth change ( r = 0.56–0.81) but not its direction , and the INT4 effect differs significantly between 650M and 3B (+0.0101, p = 0.0007) with no monotone trend to extrapolate. Third, quantizing a large model is dominated by using a small one : of eighteen scale/configuration combinations only three are Pareto-optimal over accuracy, memory and speed, and all three are 650M. The one catastrophic failure we observe is a defect of default symmetric activation scaling, not of W8A8 itself: asymmetric activation quantization, a one-line change needing no calibration, removes every damaged assay. We also give a label-free screen for at-risk targets, and report four measurement artifacts encountered during this study, three of which inverted the result they were meant to measure.
- 16 July 2026
- Published by American Chemical Society (ACS) in The Journal of Physical Chemistry B
T253X Mutations Alter Local Structure and Trigger Tyr236 Hydration in p53 DNA-Binding Domain
Mutations in the DNA-binding domain (DBD) of the p53 tumor suppressor are common in human cancers and frequently compromise its structural integrity or DNA-binding activity. Threonine 253 (Thr253) lies within the S9 aggregation-prone region, and multiple clinically reported substitutions at this position have been observed in patients with inherited cancer predisposition. However, the structural consequences of these mutations remain unclear. Here, we investigate the conformational dynamics of the wild-type (WT) p53 DBD and four T253 variants (T253A, T253P, T253I, and T253N) using all-atom molecular dynamics simulations. These simulations show that the T253 mutations remodel a local stability element centered on Tyr236, residue 253, and the S6-S7/S9-S10 loop ensemble. All T253 variants abolish or weaken the Tyr236-Thr253 hydrogen bond in the hydrophobic core. This disruption converts Tyr236 from a dry residue in WT to one that forms hydrogen bonds with water molecules, representing a common local response to mutations at position 253. Our results identify the Tyr236-residue 253 region as a local stability hotspot where buried polarity and steric bracing regulate access to an underlying β-sheet. They further provide a mechanistic basis for interpreting nonhotspot T253 variants that preserve the global fold yet subtly rewire buried interactions and hydration in a functionally critical segment.
- 16 July 2026
- Published by American Chemical Society (ACS) in The Journal of Physical Chemistry B
T253X Mutations Alter Local Structure and Trigger Tyr236 Hydration in p53 DNA-Binding Domain
Mutations in the DNA-binding domain (DBD) of the p53 tumor suppressor are common in human cancers and frequently compromise its structural integrity or DNA-binding activity. Threonine 253 (Thr253) lies within the S9 aggregation-prone region, and multiple clinically reported substitutions at this position have been observed in patients with inherited cancer predisposition. However, the structural consequences of these mutations remain unclear. Here, we investigate the conformational dynamics of the wild-type (WT) p53 DBD and four T253 variants (T253A, T253P, T253I, and T253N) using all-atom molecular dynamics simulations. These simulations show that the T253 mutations remodel a local stability element centered on Tyr236, residue 253, and the S6-S7/S9-S10 loop ensemble. All T253 variants abolish or weaken the Tyr236-Thr253 hydrogen bond in the hydrophobic core. This disruption converts Tyr236 from a dry residue in WT to one that forms hydrogen bonds with water molecules, representing a common local response to mutations at position 253. Our results identify the Tyr236-residue 253 region as a local stability hotspot where buried polarity and steric bracing regulate access to an underlying β-sheet. They further provide a mechanistic basis for interpreting nonhotspot T253 variants that preserve the global fold yet subtly rewire buried interactions and hydration in a functionally critical segment.
- Open Access
- 2 June 2026
- Published by Royal Society of Chemistry (RSC) in Physical Chemistry Chemical Physics
- Vol. 28 (25) , 15215-15221
First water-to-ligand substitution dictates Co 2+ /Ni 2+ extraction selectivity
First water-to-ligand substitution step exhibits highest free energy gap and controls solvent extraction selectivity.
- 2 June 2026
- Published by Royal Society of Chemistry (RSC) in Physical Chemistry Chemical Physics
- Vol. 28 (25) , 15215-15221
Open AccessFirst water-to-ligand substitution dictates Co 2+ /Ni 2+ extraction selectivity
First water-to-ligand substitution step exhibits highest free energy gap and controls solvent extraction selectivity.
- Open Access
DPLM: Dynamics-aware Protein Language Model via Contrastive Learning between Sequence and Molecular Dynamics Simulation Trajectory
Protein dynamics captures the interactions between residues and the conformational changes they undergo. Encoding such information could enhance a deep learning model’s ability to predict protein properties. Here we introduce a dynamics-aware protein language model (DPLM) that encodes protein dynamics into a sequence-derived representation. We represent each molecular dynamics (MD) trajectory as a sequence of backbone Cα contact maps and encode it with a Video Vision Transformer (ViViT). We then align the ViViT-derived dynamics embedding with the ESM2-derived sequence embedding using dual-view contrastive learning. The root-mean-square fluctuation correlation and dynamic cross-correlation matrix reconstruction confirm that contrastive learning incorporates dynamics information into the sequence embedding. We applied DPLM as a protein representation backbone for predicting diverse dynamics-related downstream tasks, including liquid-liquid phase separation, cryptic-site detection, and mutation effect on protein thermal stability. DPLM shows strong performance, especially where the evolutionary or structural signal is weak. These results show that learning the geometry of molecular motion yields transferable dynamic information that complements the evolutionary and structural features of existing PLMs. Significance Statement: Proteins perform their functions in constant motion. Most protein language models (PLMs) miss this dynamic dimension, learning only from sequence or fixed structure. A few emerging PLMs model dynamics through conformational snapshots or derived dynamic properties rather than the full time-resolved motion. As a result, the temporal ordering inherent in molecular motion is not explicitly captured. We introduce DPLM, a dynamics-aware protein language model that uses self-supervised learning to align protein sequences with their molecular dynamics simulation trajectories. The resulting sequence-based representation encodes general dynamic information and benefits diverse downstream tasks, especially where evolution and structure alone fall short. DPLM offers a general route to physics-informed protein representation.
Open AccessDPLM: Dynamics-aware Protein Language Model via Contrastive Learning between Sequence and Molecular Dynamics Simulation Trajectory
Protein dynamics captures the interactions between residues and the conformational changes they undergo. Encoding such information could enhance a deep learning model’s ability to predict protein properties. Here we introduce a dynamics-aware protein language model (DPLM) that encodes protein dynamics into a sequence-derived representation. We represent each molecular dynamics (MD) trajectory as a sequence of backbone Cα contact maps and encode it with a Video Vision Transformer (ViViT). We then align the ViViT-derived dynamics embedding with the ESM2-derived sequence embedding using dual-view contrastive learning. The root-mean-square fluctuation correlation and dynamic cross-correlation matrix reconstruction confirm that contrastive learning incorporates dynamics information into the sequence embedding. We applied DPLM as a protein representation backbone for predicting diverse dynamics-related downstream tasks, including liquid-liquid phase separation, cryptic-site detection, and mutation effect on protein thermal stability. DPLM shows strong performance, especially where the evolutionary or structural signal is weak. These results show that learning the geometry of molecular motion yields transferable dynamic information that complements the evolutionary and structural features of existing PLMs. Significance Statement: Proteins perform their functions in constant motion. Most protein language models (PLMs) miss this dynamic dimension, learning only from sequence or fixed structure. A few emerging PLMs model dynamics through conformational snapshots or derived dynamic properties rather than the full time-resolved motion. As a result, the temporal ordering inherent in molecular motion is not explicitly captured. We introduce DPLM, a dynamics-aware protein language model that uses self-supervised learning to align protein sequences with their molecular dynamics simulation trajectories. The resulting sequence-based representation encodes general dynamic information and benefits diverse downstream tasks, especially where evolution and structure alone fall short. DPLM offers a general route to physics-informed protein representation.
- Open Access
- 4 March 2026
- Published by Royal Society of Chemistry (RSC) in Journal of Materials Chemistry B
- Vol. 14 (13) , 4059-4069
Discovering naturally occurring antifreeze peptides from microbiome by integrating protein language models and molecular dynamics simulation
Antifreeze peptides inhibit ice crystal growth and recrystallization, and are promising components of cryoprotective formulations for cell, tissue, and food preservation, as well as anti-icing surface coatings.
- 4 March 2026
- Published by Royal Society of Chemistry (RSC) in Journal of Materials Chemistry B
- Vol. 14 (13) , 4059-4069
Open AccessDiscovering naturally occurring antifreeze peptides from microbiome by integrating protein language models and molecular dynamics simulation
Antifreeze peptides inhibit ice crystal growth and recrystallization, and are promising components of cryoprotective formulations for cell, tissue, and food preservation, as well as anti-icing surface coatings.
- Open Access
- 24 December 2025
- Published by Royal Society of Chemistry (RSC) in Physical Chemistry Chemical Physics
- Vol. 28 (2) , 1947-1962
Effect of protective mutation on structure and dynamics of APOE: a molecular dynamics simulation study
Protective mutations modulate APOE functional domain flexibility and exposure to limit aggregation.
- 24 December 2025
- Published by Royal Society of Chemistry (RSC) in Physical Chemistry Chemical Physics
- Vol. 28 (2) , 1947-1962
Open AccessEffect of protective mutation on structure and dynamics of APOE: a molecular dynamics simulation study
Protective mutations modulate APOE functional domain flexibility and exposure to limit aggregation.
- 5 December 2025
- Published by Royal Society of Chemistry (RSC) in Physical Chemistry Chemical Physics
- Vol. 27 (47) , 25498-25511
Understanding CO 2 adsorption on the surfaces of SrO and its hydroxylated variants Sr(OH) 2 · n H 2 O ( n = 0, 1, 8)
Density functional theory reveals CO 2 adsorption pathways on SrO, and Sr(OH) 2 · n H 2 O ( n = 0, 1, 8) surfaces. CO 2 forms bicarbonate (HCO 3 − ) then carbonate (CO 3 ² − ) via electron/proton transfer.
- 5 December 2025
- Published by Royal Society of Chemistry (RSC) in Physical Chemistry Chemical Physics
- Vol. 27 (47) , 25498-25511
Understanding CO 2 adsorption on the surfaces of SrO and its hydroxylated variants Sr(OH) 2 · n H 2 O ( n = 0, 1, 8)
Density functional theory reveals CO 2 adsorption pathways on SrO, and Sr(OH) 2 · n H 2 O ( n = 0, 1, 8) surfaces. CO 2 forms bicarbonate (HCO 3 − ) then carbonate (CO 3 ² − ) via electron/proton transfer.
- Open Access
- 1 December 2025
- Published by KEAI Publishing in Supramolecular Materials
- Vol. 4
Fundamental and Application of Co-assembly of Peptides and Proteins: Experiment and Computation
- 1 December 2025
- Published by KEAI Publishing in Supramolecular Materials
- Vol. 4
Open AccessFundamental and Application of Co-assembly of Peptides and Proteins: Experiment and Computation
- Open Access
- 26 November 2025
- Published by American Chemical Society (ACS) in ChemRxiv
First Water-to-Ligand Substitution Dictates Co2+/Ni2+ Extraction Selectivity
Solvent extraction is a crucial technology in the hydrometallurgical recovery of critical metals. A molecular-level understanding of how extractant ligands mediate metal transfer at the aqueous–organic interface, and how selectivity arises, is essential for elucidating the separation process. In this work, we focus on the interfacial solvation exchange during the extraction of Co2+ and Ni2+, using two representative organophosphorus extractants, Cyanex 272 and D2EHPA. Interfacial ligand substitution in the primary solvation shell, in which coordinated water molecules are replaced by extractant ligands, is the key mechanistic event governing the extraction process. Using well-tempered metadynamics simulations, we mapped the free-energy landscape of the stepwise ligand exchange at the aqueous–organic interface. The substitution of water molecules by extractant ligands is thermodynamically favorable, with the maximum free- energy differences arising in the initial step of breaking the metal hydration shell. We show that the free-energy cost (∆F ) of the first ligand-substitution step, and the corresponding energy-gap differences (∆∆F ), dictate the extraction selectivity, with Co2+ being more easily extracted than Ni2+, and Cyanex 272 exhibiting higher selectivity than D2EHPA. These results provide molecular-level thermodynamic insights into solvent extraction and selectivity.
- 26 November 2025
- Published by American Chemical Society (ACS) in ChemRxiv
Open AccessFirst Water-to-Ligand Substitution Dictates Co2+/Ni2+ Extraction Selectivity
Solvent extraction is a crucial technology in the hydrometallurgical recovery of critical metals. A molecular-level understanding of how extractant ligands mediate metal transfer at the aqueous–organic interface, and how selectivity arises, is essential for elucidating the separation process. In this work, we focus on the interfacial solvation exchange during the extraction of Co2+ and Ni2+, using two representative organophosphorus extractants, Cyanex 272 and D2EHPA. Interfacial ligand substitution in the primary solvation shell, in which coordinated water molecules are replaced by extractant ligands, is the key mechanistic event governing the extraction process. Using well-tempered metadynamics simulations, we mapped the free-energy landscape of the stepwise ligand exchange at the aqueous–organic interface. The substitution of water molecules by extractant ligands is thermodynamically favorable, with the maximum free- energy differences arising in the initial step of breaking the metal hydration shell. We show that the free-energy cost (∆F ) of the first ligand-substitution step, and the corresponding energy-gap differences (∆∆F ), dictate the extraction selectivity, with Co2+ being more easily extracted than Ni2+, and Cyanex 272 exhibiting higher selectivity than D2EHPA. These results provide molecular-level thermodynamic insights into solvent extraction and selectivity.
- Open Access
- 13 August 2025
- Published by American Chemical Society (ACS) in ACS Omega
- Vol. 10 (33) , 38112-38121
Tuning the Upper Critical Solution Temperature of the Polystyrene-Terpineol System toward a Low-Temperature Membrane Formation
The upper critical solution temperature (UCST) of binary polymer-solvent blends can create porous structures for many applications, including filtration. This study investigates how UCST can be tuned in a non-aqueous system of polystyrene and terpineol. The addition of small molecules, γ-valerolactone, oleic acid, and limonene, was tested to modify the terpineol-polystyrene phase separation temperature. The resulting porous structures were studied to determine the pore size, porosity, water flux, and rejection of test molecules to better understand the impact of the additives on both the processing temperature and porous materials' properties. The study found that the hydrogen bond propensity and miscibility of the additives significantly affected the UCST. A change of over 35 °C was observed when the additive concentration varied from 0 to 15 wt %, and the transition temperature increased or decreased depending on the additive solvent strength. The surface pore diameter was significantly altered, but the bulk pore diameter remained similar in all cases with the exception of oleic acid. The addition of small-molecule addives offers a way to control the UCST by tunning addities' solubilities and hydrogen bonding abilities in similar blends, toward lower energy use porous film synthesis.
- 13 August 2025
- Published by American Chemical Society (ACS) in ACS Omega
- Vol. 10 (33) , 38112-38121
Open AccessTuning the Upper Critical Solution Temperature of the Polystyrene-Terpineol System toward a Low-Temperature Membrane Formation
The upper critical solution temperature (UCST) of binary polymer-solvent blends can create porous structures for many applications, including filtration. This study investigates how UCST can be tuned in a non-aqueous system of polystyrene and terpineol. The addition of small molecules, γ-valerolactone, oleic acid, and limonene, was tested to modify the terpineol-polystyrene phase separation temperature. The resulting porous structures were studied to determine the pore size, porosity, water flux, and rejection of test molecules to better understand the impact of the additives on both the processing temperature and porous materials' properties. The study found that the hydrogen bond propensity and miscibility of the additives significantly affected the UCST. A change of over 35 °C was observed when the additive concentration varied from 0 to 15 wt %, and the transition temperature increased or decreased depending on the additive solvent strength. The surface pore diameter was significantly altered, but the bulk pore diameter remained similar in all cases with the exception of oleic acid. The addition of small-molecule addives offers a way to control the UCST by tunning addities' solubilities and hydrogen bonding abilities in similar blends, toward lower energy use porous film synthesis.