About

Tianyi Li is an Assistant Professor in the Department of Computer and Information Technology at Purdue University. Her research focuses on human-computer interaction (HCI) systems, where she creates AI-infused, community-centered interfaces that empower human–AI teams to address domain-specific decisions. Her work integrates design, experimental, and computational methods to understand users, build collective intelligence through crowdsourcing, and facilitate and evaluate human-AI interactions. She holds a Ph.D. in Computer Science from Virginia Tech, earned in 2020, and a B.Eng. in Computer Science from The University of Hong Kong, earned in 2015. Tianyi Li is actively seeking students interested in the intersection of HCI, computing education, and intelligent user interfaces, encouraging those from diverse backgrounds such as computer science, psychology, engineering, and related disciplines to join her research group.

Research topics

• Cell biology
• Materials science
• Biology
• Chemistry
• Immunology
• Nanotechnology
• Genetics
• Internal medicine
• Organic chemistry
• Biochemistry
• Molecular biology
• Biophysics
• Medicine
• Pharmacology
• Computational biology

Selected publications

• Multiscale Dissolution Simulation of Particles from a Tablet in Dissolution Apparatus by Coupling Discrete Element with Lattice Boltzmann Methods

  Pharmaceutical Research · 2026-05-20

  articleOpen accessSenior authorCorresponding

  PURPOSE: Dissolution kinetics of tablets is pivotal to therapeutic performance, quality assessment, and troubleshooting in manufacturing. We aimed to develop and validate a simulation framework to predict tablet dissolution kinetics under realistic hydrodynamic conditions, thereby supporting process development and quality control. METHODS: We modeled the dissolution of drug particles packed within a tablet by coupling the lattice Boltzmann method (LBM) for fluid flow and transport with the discrete element method (DEM) for particle mechanics in USP Apparatus II. The framework was first validated against experimental measurements of single‑particle dissolution, then scaled to bulk-particle (tablet‑level) simulations to assess model extensibility. RESULTS: The coupled LBM-DEM framework reproduced experimentally observed hydrodynamics and particle dynamics in a mechanically agitated dissolution device and scaled from single‑particle to tablet‑level simulations without sacrificing fidelity. CONCLUSIONS: This physics‑based framework enables the prediction of tablet dissolution under compendial hydrodynamic conditions and provides a foundation for incorporating additional particle‑mechanical phenomena (e.g., swelling and breakage) to fully model tablet disintegration and dissolution.

  PublisherOA PDFDOI

• Challenges and innovations in long-acting injectable formulations: can formulation design space be rationalized?

  Journal of Pharmacy and Pharmacology · 2025-09-26

  articleSenior author

  OBJECTIVES: To date, ~70 long-acting injectable (LAI) formulations have been developed. More than half of these formulations consist of oily solutions and suspensions containing poorly water-soluble drugs. However, numerous drugs do not fall into the category of poor solubility, such as hydrophilic small molecules, nucleic acids, peptides, and proteins. These drugs are typically formulated using biodegradable poly(lactide-co-glycolide) polymers. An important question to consider is whether there are guiding principles for selecting appropriate drugs for LAI formulations. The historical advancements and challenges associated with LAI formulations were examined to identify indicators that may predict effective drug candidates for this type of delivery system. KEY FINDINGS: Several properties of drugs, including water solubility, lipophilicity, tissue permeability, half-life (t1/2), and effective dosage, were analysed in relation to the development of LAIs. This study investigated several parameters to forecast formulation success, with a focus on achieving an optimal balance between the drug's partition coefficient (logP), which reflects both water solubility and cellular permeability, and the effective dose. SUMMARY: The current overview of recent innovations and formulation considerations indicates that a systematic approach, integrating two key parameters, logP and the effective dose of a drug, may be employed for the preliminary screening of drugs that have the potential to be formulated into LAIs with a higher probability of success in clinical applications.

  PublisherDOI

• <scp>AI</scp> / <scp>ML</scp> in Translation: <scp>PhRMA</scp> Foundation Trainee Challenge Award

  Clinical and Translational Science · 2025-10-01

  articleOpen access

  The PhRMA Foundation and ASCPT's journal Clinical and Translational Science (CTS) partnered on a Challenge Award competition to recognize trainees for outstanding papers addressing artificial intelligence (AI) and machine learning (ML) in clinical and translational science (https://www.ascpt.org/Resources/ASCPT-News/View/ArticleId/28634/CTS-Call-for-Papers-PhRMA-Foundation-Trainee-Challenge-Award). This partnership between ASCPT and the PhRMA Foundation is the most recent in a long series of successful initiatives. CTS and the PhRMA Foundation have a rich history of championing trainees and early career investigators; consequently, this collaboration is a natural evolution in the partnership. In addition, this Challenge Award competition follows our successful AI/ML Special Collection, which demonstrates the wide scope and applicability of AI/ML in translational science (https://ascpt.onlinelibrary.wiley.com/doi/toc/10.1111/(ISSN)1752-8062.ai-machine-learning). The PhRMA Foundation's expert review committee selected six papers to receive a $5000 Trainee Challenge Award. These trainee first authors are future leaders tackling a timely and challenging topic. Their papers cover a range of topics on the transformative potential of AI/ML, from clinical prediction to a methodological emphasis on model construction, validation, and robustness. Each contribution represents an innovative response to the call and has significant implications for future research. Congratulations to the authors for being selected by a panel of experts in drug development, translational science, and data science following the acceptance of their manuscripts by CTS. The award-winning papers are summarized in Table 1. Grant et al. [1] applied tree-based machine learning methods to predict progression-free and overall survival of renal cell carcinoma patients. They used patient biomarkers as input features and data from more than 1800 patients. Bhat and Ramanathan [2] predicted liver steatosis and fibrosis based on biomarker features from the liver elastography data of 5494 participants. What is most interesting about their paper is Bayesian network modeling, in which inference relationships among biomarker features were derived. In their systems biology paper, Shukla et al. [3] demonstrated a novel ML approach to identify genomic variants in breast cancer by integrating multiomics and protein structure prediction. Various types of data and knowledge bases were creatively used and streamlined. In another study, Chung and Lee [4] compared the predictive performance of ML, including regression, tree-based ensemble methods, and neural networks, with that of conventional population pharmacokinetic (PK) models. They extracted PK data of several drugs from hundreds of patient records. The study demonstrated comparable results between machine learning and PK models. Sano et al. [5] reported an interesting deep learning approach to predict disease progression of type 2 diabetes by analyzing clinical data of more than 10,000 patients. The study shed light on specific biomarkers that correlate with diabetic progression over 30 years. Finally, Weng et al. [6] discussed the implementation of out-of-distribution (OOD) detection algorithms in processing biomedical data such as images, transcriptomic data, and time series observations. Finding an “outlier” in a database could facilitate the training of a machine model. We are incredibly pleased to see these six high-quality research papers selected for the PhRMA Foundation Trainee Challenge Award in AI/ML. CTS would like to thank the expert panel for evaluating all the manuscripts submitted to the competition. CTS also wants to thank the PhRMA Foundation for its visionary support of an emerging research area that holds so much potential to reshape drug development. J. A. W. is an employee of Aditum Bio and Tempero Bio. The other authors declare no conflicts of interest.

  PublisherOA PDFDOI

• Leveraging Model Master Files for Long-Acting Injectables

  Pharmaceutical Research · 2025-01-28 · 1 citations

  articleOpen access

  The U.S. Food and Drug Administration (FDA) and the Center for Research on Complex Generics (CRCG) hosted a public workshop on May 2-3, 2024, titled "Considerations and Potential Regulatory Applications for a Model Master File". The workshop aimed to discuss the application of the Model Master File (MMF) concept in regulatory submissions that contain model integrated evidence (MIE), improving model sharing, model standardization, regulatory consistency, and regulatory efficiency. On Day 1, there was a session dedicated to MMF applications for long-acting injectables (LAIs). This perspective summarizes presentations, panel discussion, and small group discussion for the potential applications of MMFs in LAI product development, including case studies and potential situations in which MMFs can support regulatory submissions. The scientific presentations discussed the application of MMFs in mechanistic physiologically based pharmacokinetic (PBPK), multiphysics simulation, and population pharmacokinetics (popPK) models, as well as the potential utility of a model-integrated bioequivalence (MI-BE) framework. Additionally, challenges and considerations of implementing MMFs for LAIs were discussed in the panel and small groups. The anticipated benefits of MMFs are recognized among model developers, industries, and regulators.

  PublisherOA PDFDOI

• Deep Learning of CYP450 Binding of Small Molecules by Quantum Information

  Journal of Chemical Information and Modeling · 2025-01-27 · 4 citations

  articleSenior authorCorresponding

  Drug-drug interaction can lead to diminished therapeutic effects or increased toxicity, posing significant risks, especially in polypharmacy, and cytochrome P450 plays an indispensable role in this interaction. Cytochrome P450, responsible for the metabolism and detoxification of most drugs, metabolizes about 90% of Food and Drug Administration-approved drugs, making early detection of potential drug-drug interactions. Over the years, in-silico modeling has become a valuable tool for predicting drug-drug interactions. Still, conventional molecular descriptors focusing on the structural properties of drugs often overlook complex electronic interactions critical for accurate predictions. To address this, we implemented the Manifold Embedding of Molecular Surface (MEMS) approach, which retains the quantum mechanical characteristics of molecules. MEMS-generated electronic attributes were embedded and featurized for deep learning using the DeepSets architecture, where our models achieved high accuracy, particularly for cytochrome P450 enzyme 1A2 (CYP1A2), with F1 scores reaching up to 0.866. This study highlights the potential of integrating detailed electronic properties with deep learning to improve predictive models for drug-drug interactions, addressing the limitations of traditional molecular descriptors and machine-learning techniques.

  PublisherDOI

• Manifold Embedding of Quantum Information as Molecule Representation to Predict Blood-Brain Barrier Permeability by Deep Learning

  Molecular Pharmaceutics · 2025-09-30

  articleSenior authorCorresponding

  Neurological disorders continue to be a leading global health challenge, with the blood-brain barrier (BBB) presenting considerable obstacles to effective drug delivery for central nervous system (CNS) therapies. Accurately predicting BBB permeability is essential for the early stages of CNS drug design. This study utilizes Manifold Embedding of Molecular Surface (MEMS) as a quantum-informed molecule representation to improve log BB prediction using deep learning models. Employing the B3DB data set, our approach achieved competitive performance, with an average RMSE of 0.49 ± 0.06, MAE of 0.38 ± 0.05, and R2 of 0.55. The ability of MEMS to authentically encode molecular interactions facilitates a more direct modeling of log BB compared to traditional descriptors. Still, as expected, model performance is influenced by the size and quality of the data, exhibiting notable variability across different B3DB groups and imbalances in the distribution of the log BB values. Additionally, although chirality significantly influences BBB permeability, the limited stereochemical data in the data set constrain its impact. Future efforts should focus on curating high-quality, stereochemically rich measurements and addressing data imbalances to train predictive models.

  PublisherDOI

• Machine Learning Prediction of Protein Adsorption on Drug-delivering Nanoparticles: A Literature Survey and Need for Future Development

  Pharmaceutical Research · 2025-11-26 · 5 citations

  articleOpen accessSenior author

  Nanoparticles (NPs), due to their small size and large surface area, have advanced their use as drug carriers for delivering various therapeutic molecules. When entering biological environments, nanoparticles typically adsorb proteins, forming a surface layer known as a protein corona that significantly affects the biological and therapeutic functions of a delivery system. Understanding and predicting protein adsorption is essential for optimizing nanoparticle design in drug delivery, diagnostics, and therapy. Machine learning and deep learning (ML/DL) offer promising methods for designing nanoparticles with specific properties, particularly given recent advancements in computation and nanoparticle analysis. This review explores ML/DL studies of nanoparticle-protein interactions and emphasizes the popularity of Random Forest (RF) and Deep Learning (DL) models in predicting protein corona compositions. RF models are highly valued for managing high-dimensional data and offering interpretability, which helps identify key NP features influencing protein adsorption. Conversely, DL excels at modeling non-linear relationships and detecting subtle interaction patterns. While most current research focuses on protein coronas, future models may also include other biocorona components. This is particularly relevant for soft materials, such as lipid nanoparticles (LNPs), which are now approved for delivering mRNA and peptide-based vaccines. Our findings underscore the need for advanced modeling techniques and high-quality, diverse experimental data to drive innovations in nanomedicine. Combining RF and DL approaches leverages their complementary strengths to overcome the challenge of limited experimental data and further improve NP designs for biomedical use.

  PublisherOA PDFDOI

• Correction: Leveraging Model Master Files for Long-Acting Injectables

  Pharmaceutical Research · 2025-02-12

  erratumOpen access
  PublisherOA PDFDOI

• Critical quality attributes of lipid nanoparticles and in vivo fate

  Asian Journal of Pharmaceutical Sciences · 2025-09-01 · 1 citations

  articleOpen accessSenior authorCorresponding

  Lipid nanoparticles (LNPs) have emerged as versatile carriers for the delivery of genetic medicines and small-molecule drugs, offering desired benefits for therapeutic applications. Optimization of the treatment efficacy of nanocarriers necessitates a thorough understanding of the connection between pharmacokinetics and physicochemical properties. This review consolidates scientific efforts to elucidate how LNP’s physicochemical attributes influence their in vivo fate, emphasizing particle size and shape, surface electric potential and ligand-binding chemistry. By examining the interplay between LNPs and biological barriers across various administration routes, this review provides insights into tailoring LNP properties for optimal delivery and reduced off-target effects. Recommendations for future research are provided to advance the study of LNP in vivo behaviors and offer a practical framework for optimizing in vivo performance through product design parameters.

  PublisherDOI

• Manifold Kernelization of Molecular Surface to Encode Quantum Information of Electronic Attributes for Machine Learning

  Journal of Chemical Theory and Computation · 2025-10-18

  article1st authorCorresponding

  A novel concept of encoding electronic quantities on a molecular surface is developed by unsupervised kernel learning. Through optimization of the hyperparameters of Spectral Mixture (SM) kernel functions in conducting Sparse Gaussian Process (SGP) regression of electronic attributes on a surface manifold, the resultant covariance matrix, or kernel, captures the mutual relationships among the electronic quantities as well as the topology of the molecular surface. As such, a kernel, coined MKMS (Manifold Kernelization of Molecular Surface), can be treated as a symmetric positive definite (SPD) matrix to represent a molecule for machine learning. A proper neural network model was implemented to utilize SPD matrices and preserve their collected Riemannian topology for predicting the molecular properties. The prediction results of two solubility data sets support promising potentials of using MKMS to encode quantum information on a molecule and enable machine learning applications.

  PublisherDOI

Recent grants

• Toward Building a Crystal Structure Prediction Framework

  NSF · $126k · 2012–2014

• Toward Building a Crystal Structure Prediction Framework

  NSF · $390k · 2010–2013

• CAREER: Towards Fundamental Understanding and Rational Control of Crystal Growth

  NSF · $496k · 2005–2010

Frequent coauthors

• Sihui Long

  Wuhan Institute of Technology

  187 shared

• Sean Parkin

  University of Kentucky

  115 shared

• Mingtao Zhang

  79 shared

• Faquan Yu

  76 shared

• Yunping Zhoujin

  Wuhan Institute of Technology

  56 shared

• Panpan Zhou

  Lanzhou University

  53 shared

• Peng‐Yu Liang

  State Key Laboratory of Applied Organic Chemistry

  32 shared

• S. Parkin

  Max Planck Institute of Microstructure Physics

  29 shared

Education

• Ph.D., Industrial and Physical Pharmacy

  Purdue University System

Awards & honors

• Polytechnic faculty members recognized for $1M+ awards in 20…
• Polytechnic research awards - July 2024
• Polytechnic research awards - June 2024