About

Chenglong Li, Ph.D., obtained his B.Sc. in chemistry and M.Sc. in physical chemistry from Beijing University in 1985 and 1988, respectively. He worked at the Institute of Biophysics at the Chinese Academy of Sciences before earning his Ph.D. in biophysics from Cornell University in 2000. Following his doctoral studies, he spent five years in San Diego, California, conducting research as a postdoctoral fellow in structural biology at the Burnham Institute for Medical Research and as a research associate in computational chemistry at the Scripps Research Institute. In 2005, he joined the Ohio State University as a tenure-track assistant professor in the Division of Medicinal Chemistry and Pharmacognosy, where he was promoted to associate professor with tenure in 2011 and full professor in 2016. In August 2016, he was appointed as the Nicholas Bodor Professor in Drug Discovery and professor of medicinal chemistry at the University of Florida. In 2026, he was named chair of the Department of Medicinal Chemistry. His scholarly interests encompass organic chemistry, biochemistry, medicinal chemistry, physical chemistry, computational chemistry, molecular biophysics, and pharmacology. His research focuses on molecular recognition, especially protein-ligand interactions, utilizing molecular simulation, synthetic chemistry, X-ray crystallography, thermodynamic measurements, cellular techniques, and in vivo models. His current projects include developing computational methods such as a novel Multiple Ligand Simultaneous Docking strategy, and designing drugs targeting pathways and enzymes related to inflammation, epigenetics, neurodegenerative diseases, and cystic fibrosis.

Research topics

• Biology
• Chemistry
• Medicine
• Biochemistry
• Information Retrieval
• Internal medicine
• Artificial Intelligence
• Computer Science
• Cell biology
• Bioinformatics
• Neuroscience
• Psychology
• Materials science
• Pharmacology
• Endocrinology
• Cancer research
• Pathology
• Organic chemistry
• Computational biology
• Combinatorial chemistry
• Nanotechnology
• Biophysics
• Data science

Selected publications

• Homomorphic Convolution Reimagined: Eliminating Rotation Bottlenecks for Practical Privacy-Preserving CNN Inference

  2026-04-21

  article1st authorCorresponding

  The increasing adoption of Machine Learning as a Service (MLaaS) raises the demand for privacy-preserving inference, especially for convolutional neural networks (CNNs) handling sensitive visual data. Homomorphic encryption (HE) enables secure inference but remains bottlenecked by costly homomorphic rotations, whose complexity in convolution layers scales with channel and kernel dimensions. Existing state-of-the-art (SOTA) frameworks mitigate but cannot eliminate this overhead, as MIMO aggregation stage still requires rotations proportional to the output channel product. We present a homomorphic convolution framework that substantially reduces rotation cost. We redesign the SISO module to reuse rotated ciphertexts, cutting complexity from f<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">w</inf>f<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">h</inf> to $2\sqrt {{f_w}{f_h}} $. More critically, for the MIMO, we propose a channel-wise packing scheme that aligns intermediate results by channel index, enabling direct rotation-free aggregation. Under 5×5 kernels, our method achieves up to 10× speedup over SOTA. On ImageNet-scale CNNs, it delivers up to 2.3× end-to-end acceleration, providing an efficient HE-based secure inference scheme for convolutional neural networks.

  PublisherDOI

• Theaflavin reduces Achilles tendon heterotopic ossification in mice through the TGF-β/Smad signaling pathway

  Biochemical Pharmacology · 2025-10-03 · 1 citations

  articleOpen access

  Heterotopic ossification (HO) affects millions of people worldwide. TGF-β/Smad signaling pathway plays an essential role in HO of the Achilles tendon. Recent studies have found that Theaflavin can regulate the TGF-β/Smad signaling pathway, suggesting Theaflavin may have a positive effect on HO. In this study, we aimed to study the effects of Theaflavin on preventing endochondral differentiation of Tendon-derived stem cells (TDSCs) and the HO process after in the achilles tendon injury model. Here, we investigated the role of Theaflavin in a mouse model of Achilles tendon heterotopic ossification. In addition, we use TDSCs to explore the molecular mechanism of Theaflavin affecting HO. Our data showed that Theaflavin can inhibit HO of the Achilles tendon and significantly reduce the volume of mature bone tissues. SOX9 and RUNX2 were decreased significantly after Theaflavin administration. Experiments showed Theaflavin inhibited TGF-β/Smad signaling pathway during chondrogenic differentiation and osteogenic differentiation of TDSCs. Surface Plasmon Resonance Assay and Molecular docking simulation showed that a direct molecular interaction between Theaflavin and TβRI (a membrane receptor of TGF-βs). Meanwhile, Cellular Thermal Shift Assay showed Theaflavin bonded and thermally stabilized TβRI significantly. Moreover, WB and IHC displayed Theaflavin also exhibited inhibitory effects on TβRI phosphorylation. In summary, Our findings demonstrated that Theaflavin inhibited HO by down-regulating the TGF-β/Smad signal pathway, and this effect may attributed to Theaflavin binded directly to TβRI and prevented its phosphorylation.

  PublisherDOI

• Figure S2 from Oral IRAK-4 Inhibitor CA-4948 Is Blood-Brain Barrier Penetrant and Has Single-Agent Activity against CNS Lymphoma and Melanoma Brain Metastases

  2025-11-25

  articleOpen access

  &lt;p&gt;Supplemental Figure 2. CA-4948 affinity for murine IRAK-4.&lt;/p&gt;

  PublisherOA PDFDOI

• Figure S5 from Oral IRAK-4 Inhibitor CA-4948 Is Blood-Brain Barrier Penetrant and Has Single-Agent Activity against CNS Lymphoma and Melanoma Brain Metastases

  2025-11-25

  articleOpen access

  &lt;p&gt;Supplemental Figure 5: Chromatogram of CA-4948.&lt;/p&gt;

  PublisherOA PDFDOI

• Identification of Risk Factors Influencing Hemorrhage Volume in Aneurysmal Subarachnoid Hemorrhage: A Multicenter Retrospective Study

  Brain and Behavior · 2025-05-01

  articleOpen access1st author

  OBJECTIVE: This multicenter retrospective study aimed to identify significant risk factors influencing hemorrhage volume in patients with aneurysmal subarachnoid hemorrhage (SAH). METHODS: A total of 891 patients diagnosed with SAH were included from multiple medical centers. Data encompassing demographic characteristics, medical history, clinical parameters at admission, and radiographic findings were collected and analyzed. Univariate and multivariate logistic regression analyses were conducted to investigate associations between various risk factors and hemorrhage volume. RESULTS: This study identifies several factors significantly associated with increased hemorrhage volume in patients with subarachnoid hemorrhage (SAH). Multivariate analysis revealed that diabetes (P = 0.022), hypertension (P = 0.047), and saccular aneurysm morphology (P = 0.008) were independent risk factors for high hemorrhage volume. Additionally, larger aneurysm size (maximum diameter: P = 0.007, neck diameter: P = 0.021) and higher systolic blood pressure after onset (P = 0.002) were also significant predictors of increased hemorrhage volume. Factors such as age (P = 0.05) and time interval to the first CT scan (P = 0.022) were found to be associated with hemorrhage volume in univariate analysis but did not maintain independent significance in multivariate regression. CONCLUSION: This study highlights key risk factors, including diabetes, hypertension, and saccular aneurysm morphology, which independently contribute to higher hemorrhage volume in SAH patients. Management strategies focusing on early detection and control of these factors may improve clinical outcomes by reducing the risk of hemorrhagic complications. While other factors such as age and time interval to the first CT scan were associated with hemorrhage volume, they did not demonstrate independent causality in the multivariate analysis, suggesting that their role in hemorrhage volume may be secondary or context-dependent.

  PublisherOA PDFDOI

• Figure S3 from Oral IRAK-4 Inhibitor CA-4948 Is Blood-Brain Barrier Penetrant and Has Single-Agent Activity against CNS Lymphoma and Melanoma Brain Metastases

  2025-11-25

  articleOpen access

  &lt;p&gt;Supplemental Figure 3. Active IRAK-4 expression in intracranial tumors.&lt;/p&gt;

  PublisherOA PDFDOI

• Figure S7 from Oral IRAK-4 Inhibitor CA-4948 Is Blood-Brain Barrier Penetrant and Has Single-Agent Activity against CNS Lymphoma and Melanoma Brain Metastases

  2025-11-25

  articleOpen access

  &lt;p&gt;Supplemental Figure 7. Pattern of pNF-κB expression in human CNS tumors.&lt;/p&gt;

  PublisherOA PDFDOI

• GatorAffinity: Boosting Protein-Ligand Binding Affinity Prediction with Large-Scale Synthetic Structural Data

  bioRxiv (Cold Spring Harbor Laboratory) · 2025-10-01 · 2 citations

  preprintOpen access

  Protein-ligand binding affinity prediction is a fundamental task in computational drug discovery. Although substantial efforts have been made to enhance prediction accuracy using data-driven approaches, progress remains limited by persistent data scarcity. The widely used PDBbind dataset, for example, contains fewer than 20,000 experimental structures with annotated binding affinities, while a vast number of affinity measurements remain underutilized due to missing structural data. Here, we investigate this untapped potential by curating more than 450,000 synthetic protein-ligand complexes annotated with Kd and Ki values using the Boltz-1 structure prediction model. Building on this unprecedented scale of synthetic data, further augmented with over 1 million synthetic complexes from the recently released SAIR database annotated with IC50 values, we develop GatorAffinity, a geometric deep learning-based scoring function pre-trained on large-scale synthetic data and fine-tuned using high-quality experimental structures from PDBbind. Extensive evaluation on a leak-proof benchmark demonstrates that GatorAffinity significantly outperforms state-of-the-art affinity prediction methods, offering superior accuracy and generalizability. Our findings show that augmenting available experimental data with synthetic complexes can effectively address the data scarcity challenge while maintaining strong predictive reliability. By releasing the pretrained GatorAffinity model and the large-scale synthetic dataset GatorAffinity-DB, we provide a scalable and reproducible foundation for affinity prediction, virtual screening, and broader structure-based drug design applications (https://github.com/AIDD-LiLab/GatorAffinity).

  PublisherOA PDFDOI

• Adversarial Semantic and Label Perturbation Attack for Pedestrian Attribute Recognition

  ArXiv.org · 2025-05-29

  preprintOpen access

  Pedestrian Attribute Recognition (PAR) is an indispensable task in human-centered research and has made great progress in recent years with the development of deep neural networks. However, the potential vulnerability and anti-interference ability have still not been fully explored. To bridge this gap, this paper proposes the first adversarial attack and defense framework for pedestrian attribute recognition. Specifically, we exploit both global- and patch-level attacks on the pedestrian images, based on the pre-trained CLIP-based PAR framework. It first divides the input pedestrian image into non-overlapping patches and embeds them into feature embeddings using a projection layer. Meanwhile, the attribute set is expanded into sentences using prompts and embedded into attribute features using a pre-trained CLIP text encoder. A multi-modal Transformer is adopted to fuse the obtained vision and text tokens, and a feed-forward network is utilized for attribute recognition. Based on the aforementioned PAR framework, we adopt the adversarial semantic and label-perturbation to generate the adversarial noise, termed ASL-PAR. We also design a semantic offset defense strategy to suppress the influence of adversarial attacks. Extensive experiments conducted on both digital domains (i.e., PETA, PA100K, MSP60K, RAPv2) and physical domains fully validated the effectiveness of our proposed adversarial attack and defense strategies for the pedestrian attribute recognition. The source code of this paper will be released on https://github.com/Event-AHU/OpenPAR.

  PublisherOA PDFDOI

• The value of dual-energy CT virtual monoenergetic imaging and metal-artifacts reduction system after metal plate correction of congenital funnel chest in child: a single center retrospective study

  Frontiers in Pediatrics · 2025-10-23

  articleOpen access1st author

  Objective This study aimed to evaluate the utility of virtual monoenergetic (MONO) imaging and a Metal-Artifact Reduction System (MARs) algorithm for enhancing image quality in dual-energy computed tomography (DECT) of children with congenital funnel chest (CFC) following minimally invasive correction with stainless steel plate implantation. Materials and methods We retrospectively analyzed children with CFC who underwent thoracoscopic correction at our institution (January 2022 to August 2023). Their postoperative non-contrast chest CT scans were evaluated. Objective metrics (CT numbers, standard deviation-SD) were measured on five image series (120 kVp-like, 50 keV, 50 keV + MARs, 110 keV, 110 keV + MARs), and the absolute CT difference (|ΔCT|) and artifact index (AI) were calculated. Subjective image quality was scored by two independent radiologists. Given the non-normal distribution of data (Shapiro–Wilk test, P &amp;lt; 0.001), non-parametric tests (Friedman test with Wilcoxon post hoc analysis) were used for comparisons. Results A total of 114 children were enrolled. Objective metrics (SD, |ΔCT|, AI) on 110 keV images were significantly lower than on 120 kVp-like and 50 keV images ( P &amp;lt; 0.001). The application of MARs further significantly improved image quality at both energy levels ( P &amp;lt; 0.005). Subjective scores [presented as median (IQR)] confirmed that the 110 keV + MARs series provided the best image quality, which was significantly superior to all other groups ( P &amp;lt; 0.001). Conclusions The high-energy VMI (110 keV) of the dual-energy CT can effectively reduce metal artifacts, and the combination of the 110 keV and MARs algorithm could further improve the image quality, which provides great assistance for the review and follow-up of children with CFC after metal plate correction.

  PublisherOA PDFDOI

Recent grants

• NIH Grant R21CA133652

  NIH · $364k · 2012

• NIH Grant R21NS071346

  NIH · $409k · 2012

• A novel STAT3-selective inhibitor for medulloblastoma therapy

  NIH · $1.7M · 2016–2021

• Role and targeting of PRMT5 in prostate cancer

  NIH · $2.4M · 2017–2023

• Small molecule in vivo probe development targeting the IL-6/STAT3 pathway for potential multiple sclerosis therapy

  NIH · $1.3M · 2015–2019

Frequent coauthors

• Yeshavanth Banasavadi‐Siddegowda

  National Institutes of Health

  150 shared

• Marat S. Pavlyukov

  Institute of Bioorganic Chemistry

  136 shared

• Jann N. Sarkaria

  136 shared

• Ryosuke Yamada

  136 shared

• Habibe Kurt

  Brown University

  136 shared

• M. I. Shakhparonov

  Institute of Bioorganic Chemistry

  136 shared

• Christopher S. Hong

  Harvard University

  136 shared

• Gaspar J. Kitange

  University of Minnesota

  136 shared

Labs

• Department of Cellular and Systems PharmacologyPI

Education

• B.S., Chemistry

  Beijing University

  1985

• M.S., Physical Chemistry

  Beijing University

  1988

• Ph.D., Biophysics

  Cornell University

  2000

Awards & honors

• First-in-Class PRMT5-targeting epigenetic drug commercializa…