Search for: All records

Recent advances in Large Language Models (LLMs) have led to remarkable progresses in medical consultation.However, existing medical LLMs overlook the essential role of Electronic Health Records (EHR) and focus primarily on diagnosis recommendation, limiting their clinical applicability. We propose DiaLLM, the first medical LLM that integrates heterogeneous EHR data into clinically grounded dialogues, enabling clinical test recommendation, result interpretation, and diagnosis prediction to better align with real-world medical practice. To construct clinically grounded dialogues from EHR, we design a Clinical Test Reference (CTR) strategy that maps each clinical code to its corresponding description and classifies test results as “normal” or “abnormal”. Additionally, DiaLLM employs a reinforcement learning framework for evidence acquisition and automated diagnosis. To handle the large action space, we introduce a reject sampling strategy to reduce redundancy and improve exploration efficiency. Furthermore, a confirmation reward and a class-sensitive diagnosis reward are designed to guide accurate diagnosis prediction.Extensive experimental results demonstrate that DiaLLM outperforms baselines in clinical test recommendation and diagnosis prediction. Our code is available at Github. 

Manipulation of polar functional groups to extend the druggability and developability space is an important approach in the current field of drug discovery. Here, we report an editing method that enables the direct insertion of anthranilyl units into inert amides to form versatile oligoamides and cyclic peptides under exceptionally mild reaction conditions. We showcase a diverse array of pharmaceuticals, natural products, and bioactive molecules involving the mentioned scaffold insertion. The synthesis of the secondary metabolites from marine-derived fungi, the expedited construction of bioactive molecules, and the assembly of functionalized peptide macrocycles through iterative insertions highlight the synthetic utility of this method. Computational tools and experimental measurements indicate that a hydrogen bond network formed by reacting and catalytic amide enables the insertion of the anthranilyl unit into a C─N bond. 

To account for hydrodynamic interactions among solvated molecules, Brownian dynamics simulations require correlated random displacements based on the Rotne-Prager Yamakawa diffusion tensor D for a system of particles. The Spectral Lanczos Decomposition Method (SLDM) computes a sequence of Krylov subspace approximations, but each step requires a dense matrix-vector product Dq with a Lanczos vector q, and the quadratic cost of computing the product by direct summation (DS) is an obstacle for large-scale simulations. This work employs the barycentric Lagrange treecode (BLTC) to reduce the cost of the matrix-vector product while introducing a controllable approximation error. Numerical experiments compare the performance of SLDM-DS and SLDM-BLTC in serial and parallel (32 core, GPU) calculations.