Kunyang Sun

Kunyang Sun

Auto-generated profile, based on publicly available information

Kunyang Sun

University of California, Berkeley

Research Focus & Metrics

Academic Career Length

5Years (2022-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
17 September 2025 in ACS Central Science
Verified
Journal Article
Open access
KSKunyang Sun
Kunyang Sun
DBDorian Bagni
Dorian Bagni
JCJoseph M. Cavanagh
Joseph M. Cavanagh
YWYingze Wang
Yingze Wang
JSJacob M. Sawyer
Jacob M. Sawyer
THTeresa Head-Gordon
Teresa Head-Gordon
Generative machine learning models for exploring chemical space have shown immense promise, but many molecules that they generate are too difficult to synthesize, making them impractical for further investigation or development. In this work, we present a novel approach by fine-tuning Meta’s Llama3 Large Language Models (LLMs) to create SynLlama, which generates full synthetic pathways made of commonly accessible building blocks and robust organic reaction templates. SynLlama explores a large synthesizable space using significantly less data and offers strong performance in both forward and bottom-up synthesis planning compared to other state-of-the-art methods. We find that SynLlama, even without training on external building blocks, can effectively generalize to unseen yet purchasable building blocks, meaning that its reconstruction capabilities extend to a broader synthesizable chemical space than those of the training data. We also demonstrate the use of SynLlama in a pharmaceutical context for synthesis planning of analog molecules and hit expansion leads for proposed inhibitors of target proteins, offering medicinal chemists a valuable tool for discovery.

23 December 2022 in JMIR Public Health and Surveillance
Verified
Journal Article
Open access
TMTimothy Ken Mackey
Timothy Ken Mackey
AJAlan K Jarmusch
Alan K Jarmusch
QXQing Xu
Qing Xu
KSKunyang Sun
Kunyang Sun
ALAileen Lu
Aileen Lu
PDPieter C Dorrestein
Pieter C Dorrestein
Background: Antimicrobial resistance is a significant global public health threat. However, the impact of sourcing potentially substandard and falsified antibiotics via the internet remains understudied, particularly in the context of access to and quality of common antibiotics. In response, this study conducted a multifactor quality and safety analysis of antibiotics sold and purchased via online pharmacies that did not require a prescription. Objective: The aim of this paper is to identify and characterize “no prescription” online pharmacies selling 5 common antibiotics and to assess the quality characteristics of samples through controlled test buys. Methods: We first used structured search queries associated with the international nonproprietary names of amoxicillin, azithromycin, amoxicillin and clavulanic acid, cephalexin, and ciprofloxacin to detect and characterize online pharmacies offering the sale of antibiotics without a prescription. Next, we conducted controlled test buys of antibiotics and conducted a visual inspection of packaging and contents for risk evaluation. Antibiotics were then analyzed using untargeted mass spectrometry (MS). MS data were used to determine if the claimed active pharmaceutical ingredient was present, and molecular networking was used to analyze MS data to detect drug analogs as well as possible adulterants and contaminants. Results: A total of 109 unique websites were identified that actively advertised direct-to-consumer sale of antibiotics without a prescription. From these websites, we successfully placed 27 orders, received 11 packages, and collected 1373 antibiotic product samples. Visual inspection resulted in all product packaging consisting of pill packs or blister packs and some concerning indicators of potential poor quality, falsification, and improper dispensing. Though all samples had the presence of stated active pharmaceutical ingredient, molecular networking revealed a number of drug analogs of unknown identity, as well as known impurities and contaminants. Conclusions: Our study used a multifactor approach, including web surveillance, test purchasing, and analytical chemistry, to assess risk factors associated with purchasing antibiotics online. Results provide evidence of possible safety risks, including substandard packaging and shipment, falsification of product information and markings, detection of undeclared chemicals, high variability of quality across samples, and payment for orders being defrauded. Beyond immediate patient safety risks, these falsified and substandard products could exacerbate the ongoing public health threat of antimicrobial resistance by circulating substandard product to patients.

6 June 2024 in Journal of Chemical Information and Modeling
Verified
Journal Article
Open access
JLJie Li
Jie Li
OZOufan Zhang
Oufan Zhang
KSKunyang Sun
Kunyang Sun
YWYingze Wang
Yingze Wang
XGXingyi Guan
Xingyi Guan
THTeresa Head-Gordon
Teresa Head-Gordon
Determining the viability of a new drug molecule is a time- and resource-intensive task that makes computer-aided assessments a vital approach to rapid drug discovery. Here we develop a machine learning algorithm, iMiner, that generates novel inhibitor molecules for target proteins by combining deep reinforcement learning with real-time 3D molecular docking using AutoDock Vina, thereby simultaneously creating chemical novelty while constraining molecules for shape and molecular compatibility with target active sites. Moreover, through the use of various types of reward functions, we have introduced novelty in generative tasks for new molecules such as chemical similarity to a target ligand, molecules grown from known protein bound fragments, and creation of molecules that enforce interactions with target residues in the protein active site. The iMiner algorithm is embedded in a composite workflow that filters out Pan-assay interference compounds, Lipinski rule violations, uncommon structures in medicinal chemistry, and poor synthetic accessibility with options for cross-validation against other docking scoring functions and automation of a molecular dynamics simulation to measure pose stability. We also allow users to define a set of rules for the structures they would like to exclude during the training process and postfiltering steps. Because our approach relies only on the structure of the target protein, iMiner can be easily adapted for the future development of other inhibitors or small molecule therapeutics of any target protein.

25 November 2025 in bioRxiv
Verified
Preprint
Open access
ZLZi Hao Liu
Zi Hao Liu
OZOufan Zhang
Oufan Zhang
SCStefano De Castro
Stefano De Castro
KSKunyang Sun
Kunyang Sun
HGHamidreza Ghafouri
Hamidreza Ghafouri
JFJulie Forman-Kay
Julie Forman-Kay
More than two thirds of proteins in the human proteome are predicted to contain intrinsically disordered regions (IDRs), which lack stable folded structure. IDRs are critical for biological regulation and organization, as targets for post-translational modifications, and as mediators of biomolecular condensates. To address the pressing need for better structural models enabling functional insight, we developed AlphaFlex to model fully atomistic conformer ensembles for proteins predicted to have IDRs, modeled in the context of AlphaFold folded domains and an implicit bilayer for transmembrane proteins. The AlphaFlex resource provides conformational ensembles of human proteins from the AlphaFold database with identified IDRs in the Protein Ensemble Database that is mirrored in UniProt. This transformative resource of AlphaFlex ensembles provides physically and biologically relevant full-length models for IDR proteins, including scaffold proteins, those with IDR:folded-domain interactions, regulatory and condensate proteins requiring exposed binding elements, conditionally folding IDRs, and transmembrane proteins containing IDRs.

30 August 2024 in Science Advances
Verified
Journal Article
Open access
MDMatthew Du
Matthew Du
JPJuan B. Pérez-Sánchez
Juan B. Pérez-Sánchez
JCJorge A. Campos-Gonzalez-Angulo
Jorge A. Campos-Gonzalez-Angulo
AKArghadip Koner
Arghadip Koner
FMFederico Mellini
Federico Mellini
JYJoel Yuen-Zhou
Joel Yuen-Zhou
Topological insulators are insulators in the bulk but feature chiral energy propagation along the boundary. This property is topological in nature and therefore robust to disorder. Originally discovered in electronic materials, topologically protected boundary transport has since been observed in many other physical systems. Thus, it is natural to ask whether this phenomenon finds relevance in a broader context. We choreograph a dance in which a group of humans, arranged on a square grid, behave as a topological insulator. The dance features unidirectional flow of movement through dancers on the lattice edge. This effect persists when people are removed from the dance floor. Our work extends the applicability of wave physics to dance.

10 April 2026 in bioRxiv
Verified
Preprint
Open access
PDF file
KSKunyang Sun
Kunyang Sun
THTeresa Head-Gordon
Teresa Head-Gordon
Protein kinases are critical drug targets, requiring therapeutics that can modulate their active and inactive conformational states. While cofolding models can generate global folds directly from kinase sequences and ligand SMILES strings, these models have not yet been tested on their ability to recover ligand induced-fit conformational states of the kinase proteins. Here, we introduce KinConfBench, a curated benchmark of 2,225 high-quality human kinase chains to evaluate the ability of three state-of-the-art cofolding models---Boltz-2, Chai-1, and Protenix---to recover both canonical and rare conformational states. We show that geometric success metrics of a ligand pose in the active site does not correlate strongly with the correct kinase conformational state, motivating a new set of dynamical benchmarks for assessing cofolding models. While all three cofolding models achieve ~65-75% prediction accuracy for kinase conformational classification, they exhibit severe mode collapse when performing multiple inferences, show negligible structural diversity in sampling induced-fit motions, and display a prevalent ``apo-drift'' in which all three cofolding models predominately predict the kinase to be in its ligand-free state. Our results highlight that capturing ligand-induced protein conformational diversity, not just geometric fit, is critical for next-generation structure-based drug discovery.