The field of molecular dynamics and multiscale modeling is experiencing significant advancements with the integration of machine learning techniques. Researchers are developing innovative methods to bridge the gap between different time and length scales, enabling more accurate modeling of complex phenomena. One notable direction is the use of machine learning-driven workflows to orchestrate thousands of simulations operating at different scales. These workflows have the potential to revolutionize the field by enabling the simulation of complex systems that were previously inaccessible. Another area of innovation is the development of new methods for learning collective variables, which are essential for enhanced sampling techniques. These methods have shown promise in capturing the slow dynamic behavior of systems, leading to more accurate sampling and simulation results. Furthermore, the application of machine learning to drug discovery and development is also gaining traction, with new frameworks being proposed to identify drug mechanisms of action based on time-lapsed images. Additionally, automation of molecular dynamics simulations using large language models is becoming increasingly popular, reducing setup time and minimizing manual errors. Noteworthy papers include:

• A novel framework that learns collective variables directly from time-lagged conditions of a generative model, demonstrating equal or superior performance compared to existing methods.
• A molecule-auxiliary CLIP framework that combines microscopic cell video- and molecule-modalities to identify drug mechanisms of action, achieving significant improvements in drug identification and MoA recognition.
• An automated pipeline that leverages large language models to streamline the generation of MD input files, reducing setup time and offering a scalable solution for handling multiple protein systems in parallel.