Development of a parallel algorithm for the efficient compression of noisy electron cryo-microscopy micrographs of biological specimens

Imaging biological samples with an electron cryo-microscope (cryoEM) produces datasets that can be as much as 10 times bigger than an average desktop computer’s whole disk. General-purpose digital data compression can reduce the required storage space at the expense of processing time: files need to be compressed before they can be stored, and decompressed before they can be used again. The compression process involves matching character sequences, and this is something that can be sped up dramatically by splitting data in as many sections as processing units are available. While a consumer-grade computer’s Central Processing Unit (CPU) will offer a handful, its graphics hardware (GPU) will likely provide hundreds of them – thousands in the case of high-end gaming graphics cards. In the past decade, running general-purpose algorithm on graphics processing units (gpGPU) has gone from specialist development to mainstream solution. Particularly in the cryoEM field, motion correction, particle picking, CTF correction and image reconstruction can now be done in parallel on NVidia graphics cards; as these machines have become ubiquitous in cryoEM laboratories, the move to a GPU-accelerated compressed image and movie format seems only natural.

A simple profiling exercise reveals how much processing time is spent on string matching – a basic exercise of finding the longest, most repeated fragments of data so they can be substituted by much shorter codes, effectively reducing the size of the file. As different processing units (Cuda cores, in Nvidia’s terminology) can look at different parts of a micrograph, it is easy to see how having many cores will get the job done in less time.

The main body of work in the project will involve investigating ways of doing string matching in a GPU-specific way, while potentially going beyond what general-purpose compressors (e.g. zip, gzip, bzip2, rar) can achieve with images as noisy as those produced by cryoEM microscopes.

The resulting code will be distributed by the Collaborative Computational Project in Electron cryoMicroscopy (CCP-EM), who will also provide access to GitLab, a professional software repository. Additionally, the developed image and movie formats may be adopted by EMPIAR, the cryoEM dataset archive, who currently hold more than 100TB worth of data.

All research students follow our innovative Doctoral Training in Chemistry (iDTC): cohort-based training to support the development of scientific, transferable and employability skills. All research students take the core training package which provides both a grounding in the skills required for their research, and transferable skills to enhance employability opportunities following graduation. Core training is progressive and takes place at appropriate points throughout a student’s higher degree programme, with the majority of training taking place in Year 1. In conjunction with the Core training, students, in consultation with their supervisor(s), select training related to the area of their research.

Depending on the successful candidate’s qualifications, taking an ‘Introduction to Python programming’ course might be desirable. This runs regularly in the Department of Chemistry at the University of York, and familiarises the students with programmatic access to public databases, which is one the features we expect to incorporate to our software. Also, should the candidate not come from a structural biology background, a list of relevant modules from our undergraduate courses will be identified in order to bring the candidate up to speed.

The project will also provide externally-funded opportunities for teaching and training in specialised structural biology workshops in the UK and overseas through the Collaborative Computational Projects for macromolecular crystallography (CCP4) and electron cryo-microscopy (CCP-EM).

The Department of Chemistry holds an Athena SWAN Gold Award and is committed to supporting equality and diversity for all staff and students. The Department strives to provide a working environment which allows all staff and students to contribute fully, to flourish, and to excel: https://www.york.ac.uk/chemistry/ed/. This PhD project is available to study full-time or part-time (50%).

This PhD will formally start on 1 October 2019. Induction activities will start on 30 September.