Dr. Wentao Guo

Dr. Wentao Guo

She/Her/Hers

Less is more.

Talks and presentations

Physics + AI-Guided Design for Cell-Permeable Cyclic Peptide Discovery Permalink

September 16, 2025

talk, Boulder Peptide Symposium, Pennsylvania, US

Cyclic peptidomimetics occupy the sweet spot between small molecules and biologics, combining high target affinity with proteolytic stability and intrinsic cell penetration. However, their vast chemical space presents significant challenges for discovery. To address this, we (Atombeat Inc.) introduce RiDYMO.PepTx, a physics- and AI-guided platform that enables rational in silica design within an in silico cyclc peptide library constructed from more than 1000 natural and unatural amino acids. The platform integrates drug-likeness and permeability neural networks to prioritize candidates for rapid synthesis and testing.

Predictive and Generative Artificial intelligence towards Polymers Permalink

August 19, 2025

talk, ACS, Washington, D.C, US

Understanding the three-dimensional conformation of polymers is essential for connecting molecular structure to macroscopic material properties. However, generating reliable polymer conformations remains a major challenge due to their structural flexibility, diversity, and limited availability of high-quality reference data. Building on our initial work predicting polymer properties with its monomer structure, we present our following work, PolyConf, a generative modeling framework that predicts polymer conformations directly from molecular graphs.

Enhancing AI-aided drug design with Uni-QSAR (on behalf of Dr. Cui) Permalink

August 19, 2025

talk, ACS, Washington, D.C, US

The application of AI/ML in drug discovery faces critical challenges in data scarcity scenarios, particularly in deriving robust quantitative structure-activity relationships (QSAR) with limited molecular annotations. We present Uni-QSAR, an automated framework that overcomes these limitations through multidimensional molecular representation learning. Our approach innovatively integrates 1D sequence patterns, 2D topological graphs, and 3D conformer geometry into a unified embedding space, enhanced by pretraining on large-scale unlabeled molecular data. Without manual parameter tuning, Uni-QSAR achieves state-of-the-art performance on major benchmarks: outperforming existing methods in 14/22 tasks (Therapeutic Data Commons) and 26/30 tasks (MoleculeACE). Notably, we demonstrate through Bindingnet analysis that 3D conformer-aware modeling significantly improves binding activity prediction accuracy. This fully automated system provides researchers with an efficient tool for predictive molecular profiling, substantially advancing AI-driven drug discovery pipelines by reducing dependency on labeled data while maintaining prediction robustness.

Vinyl Carbocation Chemistry: Confronting the Computational Challenges That Matter Permalink

May 29, 2025

talk, NSF center for computer assisted synthesis, Pennsylvania, US

Vinyl carbocations, a intriguing reaction intermediates are too unstable to observe in reality, existing only as highly reactive intermediates. In this projectm we use ab initio computational approaches combining machine-learning potentials to investigate its half-life time and stability in solution with explicit solvation modeled in the system.

Flexibility, π-π stacking, and ylide stabilization in organometalic catalyzation Permalink

August 18, 2024

Invited virtual talk, ACS, Colorado, US

Recent experimental work has demonstrated the potential of mixed-ligand Rh(II) paddlewheel complexes, such as Rh2(OAc)3PhTCB, to enhance cyclopropanation yields and selectivity compared to traditional catalysts with identical ligands like Rh2(OAc)4. In this computational study using density functional theory (DFT), we explore the mechanistic basis for the improved performance of Rh2(OAc)3PhTCB. The mixed-ligand design offers increased catalyst flexibility, enabling the dissociation of the PhTCB ligand from the vacant Rh site and the formation of stabilizing π-π stacking interactions with the substrate in a three-benzene sandwich structure. Along the reaction pathway, the Rh-carbene intermediate forms an explicit C-S bond, resulting in an ylide structure that represents a deep energy minimum. The unique functionality of the Rh2(OAc)3PhTCB catalyst arises from its ability to modulate the energetic landscape and promote cyclopropanation while suppressing side reactions. These computational insights highlight the power of mixed-ligand Rh(II) catalysts and provide a framework for the rational design of improved Rh paddlewheel catalysts.

Run Wild, Rh-carbenes! Permalink

March 14, 2024

Lightning talk, R. Bryan Miller Symposium, California, US

Buckle up for a wild run through the world of Rh-catalyzed CHCR reactions!

Post-transition state bifurcation in Rh-catalyzed reaction Permalink

August 04, 2022

Talk, University of California, Los Angeles, California, US

Rh-catalyzed C-H insertion reactions to form β-lactones suffer from post-transition state bifurcations, with the same transition states leading to ketones and ketenes via fragmentation in addition to β-lactones.

Near-infrared fluorescent probe labels Aβ plaques through in vivo bioorthogonal reactions

December 17, 2017

Talk, 5th National Symposium for Chem Undergrad, Peking Univeristy, Beijing, China

Near-infrared fluorescent probes are increasingly being explored as markers for Amyloid-beta (Aβ) plaques, playing a crucial role in the early detection of Alzheimer’s disease. In this study, we aim to engineer a probe that leverages bioorthogonal reactions, enabling the activation or deactivation of fluorescence through light modulation. 💡