Descripció del projecte

This project aims to develop novel multiscale molecular potentials based on artificial neural networks and apply them in the fields of structural biology, computational chemistry and drug discovery. The project will benefit from the active development of Neural Network Potentials (NNPs) at Acellera and expand them to new areas of drug discovery, like ligand docking.

Machine learning potentials can act as fast and accurate potential energy functions, effectively replacing conventional force fields. Their main advantage is that they can be trained to approximate high accuracy compute-intensive methods, enabling their accuracy orders of magnitude faster than the reference methods. Thanks to that, NNPs have found application in quantum chemistry, protein folding and molecular property prediction. Acellera has actively participated in the field, developing a fully differentiable molecular dynamics engine named TorchMD that enables a quick design and testing of NNPs as molecular force fields. The company is also involved in the OpenMM consortium, which enables fast deployment of NNPs and makes the advances accessible to a wide audience of computational chemists.

The candidate will participate in the development of a training pipeline for novel NNPs and their application in real drug discovery problems. The candidate will curate structural data sets, explore new training approaches and neural network architectures, train and validate the NNPs, and use them in prospective projects. The main focus will be on ligand docking, however the pipeline could be applied in other areas of drug discovery too. Ligand docking is known to suffer from low accuracy. It is usually applied in large scale virtual screening campaigns, as a means to increase the probability of finding a hit compound. More accurate, MD- or QM-based methods produce more reliable predictions, but are substantially more expensive. The candidate will improve docking methods with NNPs and create a next generation docking application.

Finally, the candidate will integrate their method and NNPs into ACEMD, Acellera’s molecular dynamics software, and in the platform, which contains various applications for drug discovery and molecular simulations.

During their work on this project, the researcher will have access to state-of-the-art computational resources. This project is expected to lead to publishable results in high-impact scientific journals.


Si t’interessa l’oferta, omple el pdf amb les teves dades i envia´l a