Shengchao Liu

Artificial intelligence (AI) has accelerated scientific prediction and design, but accurate predictions and promising structures alone do not explain how scientific systems behave. Using molecular science as our primary lens, we envision a progression from property prediction through structure generation toward a new paradigm coined Wave Intelligence. It integrates complementary modalities into a shared latent space and learns how those representations evolve under changing conditions. The term "Wave'" evokes the dynamic, evolving nature of scientific processes. This paper further outlines the research challenges and evaluation principles for developing Wave Intelligence and assessing its contribution to mechanistic understanding.

Current frame-based RNA generative models represent each nucleotide with a single rigid coordinate frame. However, our analysis of seven nucleotide frame constructions across 11,497 static RNA chains and 31,432 multi-state relation groups reveals a clear trade-off: base-centered frames better preserve base-mediated relations, whereas sugar-centered frames better support sugar-phosphate backbone reconstruction. Motivated by this complementarity, we introduce DuetRNA, a joint sequence-structure generation model for RNA co-design that performs SE(3) flow matching over coupled base-and-sugar frames. Empirically, we evaluate DuetRNA under both the inverse-folded (IF) backbone-design protocol and generated-sequence (GS) co-design protocol. DuetRNA improves scTM-validity over the strongest matched baselines by 7.67 and 4.33 % under IF and GS, respectively. Controlled ablations further support the necessity of the dual-frame modeling in DuetRNA. The official DuetRNA repository is https://github.com/XjunLi/DuetRNA.

A bstract: Chromosomes are the carriers of genetic information. Further understanding their 3D structure can help reveal gene-regulatory mechanisms and cellular functions. However, high-resolution 3D structures are often missing due to the high cost and inherent noise of experimental screening. A standard pipeline for reconstructing the chromosome 3D structure first applies the single-cell Hi-C high-throughput screening method to measure pairwise interactions between DNA fragments at different resolutions; then it adopts computational methods to reconstruct the 3D structures from these contacts. These include traditional numerical methods and deep learning models, which struggle with limited model expressiveness and poor generalization across resolutions. To handle this issue, we propose InertialGenome, a novel transformer-based framework for robust and resolution-agnostic chromosome reconstruction. InertialGenome first adopts the inertial frame for the pose canonicalization. Then, based on such an invariant pose, it proposes a Transformer with geometry-aware positional encoding, leveraging Nyström estimation. To verify the effectiveness of InertialGenome, we conduct experiments on two single-cell 3D reconstruction datasets with four resolutions, reaching superior performance over all four computational baselines. Additionally, we observe that the 3D structure reconstructed by InertialGenome is more in line with the results of real experimental results on two functional verification tasks. Finally, we leverage InertialGenome for cross-resolution transfer learning, yielding up to a 5% improvement from low to high resolution. The source code is available at https://github.com/yize1203/InertialGenome .

Molecular dynamics (MD) simulation is a key tool in drug discovery for predicting protein-ligand binding affinities, transport properties, and pocket dynamics. While advances in numerical and machine learning (ML) methods have improved MD efficiency, accurately modeling long-timescale dynamics remains challenging. We introduce NeuralMD, an ML surrogate that accelerates and enhances MD simulations of protein-ligand binding. NeuralMD employs a physics-informed, multi-grained, group-symmetric framework comprising (1) BindingNet, which enforces symmetry via vector frames and captures multi-level protein-ligand interactions, and (2) an augmented neural differential equation solver that learns trajectories under Newtonian mechanics. Across ten single-trajectory and three multi-trajectory tasks, NeuralMD achieves up to 15 × lower reconstruction error and 70% higher validity than existing ML baselines. The predicted oscillations closely align with ground-truth dynamics, establishing NeuralMD as a foundation for next-generation protein-ligand simulation research.

AI models have shown great potential in structure-based drug design, generating ligands with high binding affinities. However, existing models have often overlooked a crucial physical prior: Atoms must maintain a minimum pairwise distance to avoid atomic collision, a phenomenon governed by the balance of attractive and repulsive forces. To mitigate such atomic collisions, we propose NucleusDiff. It enforces spatial distance constraints between atomic nuclei and auxiliary mesh points placed on a spherical surface around each atom, approximating van der Waals boundaries to reduce atomic collisions. We quantitatively evaluate NucleusDiff using the CrossDocked2020 dataset and a COVID-19 therapeutic target, demonstrating that NucleusDiff reduces collision rate by up to 100.00% and enhances binding affinity by up to 22.16%, surpassing state-of-the-art models for structure-based drug design. We also provide qualitative analysis through manifold sampling, visually confirming the effectiveness of NucleusDiff in reducing atomic collisions and improving binding affinities.

In drug discovery, molecular dynamics (MD) simulation for protein-ligand binding provides a powerful tool for predicting binding affinities, estimating transport properties, and exploring pocket sites. There has been a long history of improving the efficiency of MD simulations through better numerical methods and, more recently, by utilizing machine learning (ML) methods. Yet, challenges remain, such as accurate modeling of extended-timescale simulations. To address this issue, we propose NeuralMD, the first ML surrogate that can facilitate numerical MD and provide accurate simulations in protein-ligand binding dynamics. We propose a principled approach that incorporates a novel physics-informed multi-grained group symmetric framework. Specifically, we propose (1) the BindingNet model that satisfies group symmetry using vector frames and captures the multi-level protein-ligand interactions, and (2) an augmented neural differential equation solver that learns the trajectory under Newtonian mechanics. For the experiment, we design ten single-trajectory and three multi-trajectory binding simulation tasks. We demonstrate the efficiency and effectiveness of NeuralMD, achieving over 1K$\times$ speedup compared to standard numerical MD simulations. NeuralMD also outperforms all other ML approaches, achieving up to 15$\times$ reduction in reconstruction error and 70% increase in validity. Additionally, we qualitatively illustrate that the oscillations in the predicted trajectories align more closely with ground-truth dynamics than those of other machine-learning methods. We believe NeuralMD paves the foundation for a new research paradigm in simulating protein-ligand dynamics.