Artificial intelligence and machine learning

Artificial Intelligence uses powerful computers to apply algorithms that mimic the capabilities of the human brain to find patterns in big datasets, through methods like machine learning, deep learning, and generative models.

Our researchers are using AI and machine learning in many areas of research, including the analysis of complex omics data sets, structural biology, image analysis, and drug discovery.

These technologies have extraordinary potential to enhance biomedical research.

Artificial Intelligence at WEHI

We are applying a variety of AI approaches to molecular data, including genomic sequencing and gene activity measurements, to better understand health and disease.

We are using AI in molecular data analysis, such as whole genome sequencing, transcriptomics, proteomics, multiomics and spatial omics.

WEHI researchers are involved in many big data projects, harnessing vast public datasets that provide opportunities for AI-driven discoveries.

Building on existing strengths in bioinformatics, computational biology, data science, machine learning, and data-generating technologies, WEHI is expanding its capacity in artificial intelligence and aims to become a leader in AI-enabled biomedical discovery in Australia.

Flagship AI projects

Some of our key areas of research using AI are:

  • Multi-omics – we are taking molecular measurements of multiple tissues from patient cohorts and integrating these data to find molecular subtypes and biomarkers of outcome and drug response. One example of this is examining immune cells in the blood at single-cell resolution to learn more about disease progression and identify biomarkers to predict patient prognosis or identify the best drug therapies.
  • Drug discovery – we are using AI to design novel medicines and accelerate the drug development pipeline.
  • Imaging-omics – biomedical images contain detail that is hidden from the human eye. We are using Deep Learning techniques to extract this information from medical images like MRIs and DEXA (dual x-ray absorptiometry), which can then be analysed using statistical techniques to reveal insights.

Imagined protein

Researchers are using AI to solve scientific problems, such as designing proteins that don’t exist in nature but could meet specific treatment needs if they did.

The biological function of a protein is largely determined by its 3D shape. Postdoctoral researchers Dr Marjan Hadian-Jazi and Dr Richard Birkinshaw first teach an AI network about the properties that existing proteins of different shapes possess, then specify a list of properties they’d like a new protein to have.

Imaging specialist Dr Lachlan Whitehead’s animation represents how the AI network starts with “noise” and gradually adjusts each parameter until it “imagines” a protein that is likely to meet those specifications and be feasible to produce.

Sign up to our quarterly newsletter Illuminate

Find out about recent discoveries, community supporters and more.

Illuminate Winter 2024
View the current issue