We've used mobile technology coupled with crowd sourcing to allow people to “score” protein crystallisation images. As more people score we get a better training set to help our machine algorithms learn how to score for us.

The challenge

Finding protein crystals

In order to figure out the molecular structure of biologically important molecules like proteins, scientists must use a method called X-ray crystallography. Protein crystals are an essential part of this method. Like many other research institutions, we are interested in the molecular structure of proteins because it is essential for our biotechnology research. But finding protein crystals takes a lot of time!

Protein crystals used in X-ray crystallography to determine protein structure.

In order to grow a protein crystal we need to set up thousands of crystallisation experiments. These consist of different mixtures of chemicals combined in small droplets with a concentrated pure protein sample. Unfortunately, there is no way of knowing which combination of chemicals will work for any particular protein, so for every protein, many different droplets are set up.

We then take photos of the experiments over many weeks and look at them to see if crystals have formed. Less than 1 per cent of the experiments will have crystals.

Our response

Cinder: just swipe right!

We developed Cinder so we can crowd source our search for experiments with crystals. The information gathered through Cinder will help us develop machine image analysis tools that can find crystals automatically, similar to the way Facebook can recognise people's faces in photos.

One of the challenges of identifying crystals is they can grow in many different ways and in many different shapes. When we think of crystals we think of a grain of salt or a diamond. But crystals can also be shaped long and thin like spaghetti, or as a cluster of tiny needles like a sea urchin. They may be big and really easy to spot, or hidden in a cloud of fog-like precipitated protein.

This means our crystal-spotting computer program can't just look for one type of crystal. It has to be trained on a large set of different images - some with crystals, some with no crystals.

Humans are very good at finding crystals in images, and we have developed Cinder so that people can help us create the training set which will be used to help the computer program learn how to spot crystals.

Cinder is an easy app to use - just swipe right for crystals.

The results

Speeding up drug development

Classification data of each experiment is sent back to CSIRO and the crystals we find will be used to generate atomic structures of proteins, which are used to help scientists understand the protein.

Atomic structures generated through X-ray crystallography have been used to understand how DNA works, to develop drugs against diseases like HIV, and to reveal the molecular basis of photosynthesis.

This technology will help speed up the rate at which we can complete discovery work for our industry partners.

If we can get machines to do the scoring it will free up millions of dollars worth of time for scientists worldwide, and provide a platform to significantly speed up the entire discovery phase of drug development.

Every swipe is helping us develop a better understanding of biology!

Do business with us to help your organisation thrive

We partner with small and large companies, government and industry in Australia and around the world.

Contact us

Your contact details

First name must be filled in

We'll need to know what you want to contact us about so we can give you an answer.