Deep Learning Models of Genetic Variant Effects on 3D Genome Organisation

• PhD

This project will develop deep learning models that predict 3D genome architecture directly from DNA sequence. Building on CNN-based sequence-to-function frameworks developed by Dr Alexander Sasse, the project will use Hi-C data from the F1 hybrid (B6/CAST) mouse system, where natural genetic variation between parental haplotypes enables causal testing of model predictions. The research aims to train and validate models across multiple cell types, then use them to predict how genetic variants alter chromatin architecture and identify sequence features that shape genome organisation.

The student will work with genomic sequence and Hi-C datasets using Python and deep learning frameworks (PyTorch/TensorFlow), working closely with collaborating laboratories and with opportunities to train in Germany with Dr Sasse. The ideal candidate will have a strong quantitative background and an interest in machine learning, genomics, and 3D genome biology.

The Allan Lab focuses on the molecular regulation of the immune system, in both health and disease. We have a long record of combining cutting-edge technologies (RNA-Seq, single cell RNA-Seq, ATAC-Seq, in situ HiC, capture HiC, Pore-C, chromosome paint, among others) with bioinformatics to make impactful discoveries.

We enjoy highly collaborative projects and have well established networks with scientists at WEHI, across Australia and around the world. This provides an excellent interdisciplinary training environment for students, as well as opportunities to build valuable networks.