We are looking for a highly motivated machine learning/AI student to join a multidisciplinary team across 2- continents aimed at improving the workflow of protein structure solution by cryo-electron microscopy (cryo-EM), a cutting-edge technology that was awarded the Nobel prize in Chemistry in 2017.
Cryo-electron microscopy is a specific technique for preparing biological samples cooled to cryogenic temperatures and embedded in an environment of vitreous water. used in transmission electron microscopy. Developed in the early 1980s, this technique reduces irradiation damage caused by the electron beam. In the past five years, it has become the goto technique for studying the structural arrangement of biological samples, sometimes achieving near atomic resolution.
The project we propose sits at the intersection between 3D structural biology of biological molecules and machine learning approaches. In Sanofi, where the technology was adopted in late 2020, cryo-EM has already contributed to more than 13 projects in 1 year and we plan to continue accelerating this progress by applying machine learning approaches to the solution of structures. We have identified a bottleneck in the data analysis workflow which consists of identifying 3D objects within a 3D volume that is the output of common software used. In some cases, the volumes do not contain enough detailed information that such tasks become exceedingly difficult.
The project will aim at identifying single components of 3D volumes using the database that is available for 3D structures as a data source (EMDB), plus our internal data. The student will work on developing an algorithm that includes the shape of commonly used drug ligands to be able to identify such shapes in new 3D volumes. Moreover he/she will also work on the fitting and the use of ensemble models or fragments to solve the molecular structure of the whole complex. The successful candidate will also be able to propose innovative ideas on the use of Machine Learning (ML) in the entire process.
The successful candidate will be exposed to a range of ML methods including but not limited to: reinforcement learning with 3D spatial reasoning, SE(3)-equivariant graph neural networks, forward generative modelling, various geometric techniques used in AlphaFold2 for iterative refinement, multiple GPU training, and incorporation of complex structural priors.
The novel method will be used to further accelerate the drug design process in Sanofi and bring medicines to patients quicker.
Interested students should apply by the 31st of December 2021 with a complete and up-todate CV and a half-page cover / cover letter. If you wish to have more information on the project, do not hesitate to contact us (details below).
Martin L. Rennie et al. Structural basis of FANCD2 deubiquitination by USP1−UAF1, Nature Structural & Molecular Biology (2021). DOI: 10.1038/s41594-021-00576-8 Simon Batzner and Albert Musaelian and Lixin Sun and Mario Geiger and Jonathan P. Mailoa and Mordechai Kornbluth and Nicola Molinari and Tess E. Smidt and Boris Kozinsky SE(3)- Equivariant Graph Neural Networks for Data-Efficient and Accurate Interatomic Potentials, (2021),Jul., Xarchiv, Review Nature Porto folio.
- Required qualifications: M1
- Location: Vitry sur Seine, Ile-de-France
- Start of the internship: January, March 2022
- Person to contact:
Chiara Rapisarda, Lab head in cryo-EM, Sanofi (email@example.com)
Elton Rexhepai (Elton.REXHEPAJ@sanofi.com)
- Application deadline: 31st December 2021
- Link to the company web site: https://www.sanofi.com/