Adapting machine learning algorithms to learn the rules of embryonic development

We are seeking a talented, creative, and friendly individual to join our interdisciplinary team. You will adapt cutting-edge methods from machine learning and numerical analysis to discover PDE-based dynamical models that describe how simple embryos develop into complex animals. This studentship is fully funded for 4 years as part of an ERC Starting Grant award.

PROJECT:

Like many other areas of science, developmental biology is in the middle of a “data revolution”. Breakthrough advances in genomics, single cell sequencing and high-resolution microscopy have enabled researchers to collect quantitative data at a scale that was previously unimaginable. This data charts the remarkable processes that allow very simple, single-celled embryos to self-organize into complex animals with elaborate patterns, shapes, and forms. However, as our descriptions of developing organisms become more quantitative, then so too must the language that we use to understand them; we must move beyond verbal models and towards predictive mathematical theories, following the path of other data-rich fields such as physics. However, unlike physics, many of the fundamental laws remain unknown. Therefore, here, we will use machine learning methods – applied to large quantitative datasets – to systematically “learn” the laws of embryonic development from scratch. Initially, we will focus on understanding how the anteroposterior (i.e., head-to-tail) pattern is set up in early embryos, using data from our close collaborators from the University of Cambridge.

APPROACH:

Dynamical models, especially PDEs, are widely used to explain how developmental patterns self-organize during development [1]. In this project, you will adapt methods from machine learning and numerical analysis to discover the PDEs that govern embryonic pattern formation. Recent advances from other fields have shown how dynamical systems can be learned de novo given sufficient observations of the dynamics [2]. We will optimize and adapt these approaches to match the types of data available in our experimental systems, using e.g., gradient-descent optimizers [3], physics-informed neural networks [4], Koopman operators [5], and advanced evolutionary algorithms [6]. Algorithms will be developed in the highly performant Julia programming language within the mature DifferentialEquations.jl ecosystem and made freely available to the research community. 

CANDIDATE BACKGROUND:

Given the sophisticated numerics involved in this project, candidates should have strong programming and quantitative skills. Some experience with scientific computing (e.g., in Julia/python/C++/MATLAB) is essential; familiarity with code management (e.g., github) is desirable. No formal training in developmental biology is required, although curiosity is essential. 

To be considered for PhD study at the University of Aberdeen, applicants must typically hold a UK Honours Degree with a 2:1 classification (or international equivalent). Applicants with a 2:2 undergraduate degree may also be considered if they hold a commendation or distinction at master’s level, or have significant relevant experience.

FURTHER INFORMATION:

The project will be based in Aberdeen in the beautiful Northeast of Scotland. Please email Tom with a CV for informal enquiries: [email protected].

Lab website: https://twhiscock.github.io/

———————————

APPLICATION PROCEDURE:

• Formal applications can be completed online: https://www.abdn.ac.uk/pgap/login.php
• You should apply for Medical Sciences (PhD) to ensure your application is passed to the correct team for processing (the programme applied for may not be representative of the programme which will be offered to a successful candidate, this is for administrative purposes only)
• Please clearly note the project title and lead supervisor in the respective fields on the application form
• Your application must include: A personal statement, an up-to-date copy of your academic CV, and clear copies of your educational certificates and transcripts.
• Pease provide two academic references with your application.
• Please note: you DO NOT need to provide a research proposal with this application
• Applications for this project may be shared with any external funders of this PhD Studentship, and any external members of the supervisory team.
• If you require any additional assistance in submitting your application or have any queries about the application process, please don’t hesitate to contact us at [email protected]