Predicting protein-ligand binding affinity with high accuracy is critical in structure-based drug discovery. While docking methods offer computational efficiency, they often lack the precision required for reliable affinity ranking. In contrast, molecular dynamics (MD)-based approaches such as MMGBSA provide more accurate binding free energy estimates but are computationally intensive, limiting their scalability. To address this trade-off, we introduce an active learning framework that automates molecule selection for docking and MD simulations, replacing manual expert-driven decisions with a data-efficient, model-guided strategy. Our approach integrates fixed - partly pre-trained deep learning - molecular embeddings (MolFormer, ChemBERTa-2, and Morgan fingerprints) with adaptive regression models (e.g. Bayesian Ridge and Random Forest) to iteratively improve binding affinity predictions. We evaluate this approach retrospectively on a new dataset of 60,000 chemically diverse compounds from ZINC-22 targeting the MCL1 protein using both AutoDock Vina and MMGBSA. Our results show that incorporating MMGBSA scores into the active learning loop significantly enhances performance, recovering 79.9% of the top 1% binders in the whole dataset, compared to only 6.7% when using docking scores alone. Notably, MMGBSA exhibits a stronger correlation with experimental binding affinities than AutoDock Vina on our dataset and enables more accurate ranking of candidate compounds in a runtime efficient way. Furthermore, we demonstrate that a one-at-a-time acquisition active learning strategy consistently outperforms traditional batched acquisition, the latter achieving just 78.4% recovery with MolFormer and Bayesian Ridge. These findings underscore the potential of integrating deep learning-based molecular representations with MD-level accuracy in an active learning framework, offering a scalable and efficient path to accelerate virtual screening and improve hit identification in drug discovery.

Small-angle scattering (SAS) is widely used in structural biology, soft matter, and colloidal science to probe molecular structures in solution. SAS rests on a single physical principle: wave interference from a distribution of scatterers, averaged over orientations. Yet the theoretical foundations of SAS are spread across the literature, often based on differing notation, definitions, and implicit assumptions. We present the theory of SAS in solution from first principles as a continuous derivation, spanning the scattering of a single electron to the observed intensity of a molecular solution and its comparison with atomistic structural models. The derivation is explicit throughout — approximations, averaging procedures, and algebraic manipulations are stated rather than assumed — and is independent of the probe (X-ray or neutron) and applicable to both rigid and flexible molecules. The framework resolves several ambiguities in the current literature, notably the role of background subtraction as a theoretical rather than a purely experimental operation and the role of boundary cross-terms in justifying that subtraction. A central result is that analytical scattering calculations and approaches based on explicit-solvent molecular dynamics, typically treated as distinct traditions, are realizations of the common theoretical framework derived here. As the precision and reproducibility of SAS data continue to increase, this unified framework provides a basis for integrating theory, simulation, and experiment in future developments of SAS.

Biological membranes display remarkable compositional diversity with functional consequences. In viral infections, lipids play active roles beyond forming a structural barrier: they modulate fusion protein binding, and tune membrane mechanics to prime them for efficient fusion, while viruses hijack host lipid metabolism to support replication. This highlights how membrane remodeling is exploited throughout the viral life cycle and points to potential avenues for antiviral intervention.