Releases: gnina/gnina
v1.3
This release updates the underlying deep learning framework to Torch, resulting in more computationally efficient docking and paving the way for seamless integration of other deep learning methods into the docking pipeline. We retrained our CNN scoring functions on the updated CrossDock2020 v1.3 dataset and introduce knowledge-distilled CNN scoring functions to facilitate high-throughput virtual screening.
v1.1
Implementation of easy covalent docking. Can specify SMARTs pattern for ligand atom and chain:resid:atomname for the receptor atom and docking will only explore conformations where these atoms form a covalent bond. OpenBabel bonding heuristics are used to determine the initial atom placement, but can be overridden by explicitly specifying ligand coordinates. The geometry of the covalent complex can be optional optimized with UFF.
Various bug fixes and updates to the build system.
v1.0.3
v1.0.2
v1.0.1
v1.0
The GNINA 1.0 Release. Includes support for CNN scoring throughout the docking pipeline, a default ensemble of CNNs that significantly outperforms Vina at scoring, convenient flexible docking, and support for custom empirical and CNN scoring functions.
The provided binary includes almost all dependencies in the most compatible manner possible. It is intended for evaluation only, not production use, as the focus on compatibility results in a reduction in performance. To use GPU acceleration, your CUDA driver must be >= 410.48.
Docker images are available at https://hub.docker.com/u/gnina