A new AI system has enabled the discovery of a novel role for ‘smell-sensing’ genes in colon cancer
Humans have around 400 olfactory ‘smell-sensing’ genes – the largest gene family in humans – that are turned on in the nose and other parts of the body, allowing us to smell at least one trillion different odours. Up until now, the role of these olfactory genes outside the nose has been largely unknown.
A recent study, published in Molecular Systems Biology, used multiple layers of artificial intelligence (AI) to identify these genes involved in the organisation of colon cancer cells. This revealed that smell-sensing genes can contribute to this cancer-associated process along with key colon cancer genes and highlighted their potential role in disease spread and prognosis.
The discovery was enabled by the development of an innovative AI system, called Knowledge-and Context-driven Machine Learning (KCML) that enables researchers to study microscopy images in greater detail to understand more about the function of genes in specific context. KCML has first been applied to colon cancer but is widely applicable in other diseases too.
The researchers used computer vision algorithm to detect changes in cell appearance and organisation. The algorithm was fed information from robotic microscopy, in collaboration with researchers from the University of Zurich, to image millions of colon cancer cells. By reducing the expression of the ‘smelling’ genes within these cells, they were able to understand more about the role they play in carcinogenesis.
Expression is when genes are activated to produce certain proteins and molecules. Researchers in this study found that reducing the expression of smell-sensing genes in colon cancer cells, a process known as perturbation, can inhibit cells from spreading, potentially by restraining the ability of cells to move. The same behaviour is also observed in the perturbation of key cancer genes.
Dr Heba Sailem, Sir Henry Wellcome Research Fellow at the Institute of Biomedical Engineering in the UK, a lead author on the study, explained: “With all this big imaging data, we have a powerful means to better understand how every single gene contributes to cancer cell behaviour. I have developed an AI system that is guided by prior knowledge of gene function that allows us to learn much more from this data than would be possible using existing methods.”
“When humans look at complex scenes, they interpret the images in light of their previous experience and visual memories (prior knowledge). However, computers just see images as a large matrix of numbers, they will not see shapes and structures. Computer vision is about training the computer to see what the human can see. Through AI, we are able to identify how turning genes off affects the characteristics,
shape and structure of cells and tissue. Usually, it is a very lengthy process for humans to interpret numbers from thousands of images, each with thousands of cells. Computer vision can achieve that in a few days,” she added.
Dr Sailem’s work has focussed on studying cells in culture, and the next step will be to link these findings through to real patient data. She is also keen to apply her AI model to study the behaviour of genes in different cancers, including prostrate, breast and lung.
WHY IT MATTERS
Colorectal cancer is the third most common cancer in the UK and the second most common cause of cancer deaths.
Professor Mark Lawler, chair in translational cancer genomics, Centre for Cancer Research and Cell Biology, Queen’s University Belfast and Bowel Cancer UK medical advisor, welcomed the application of the new AI model in colorectal cancer, commenting the study showed the “power of data in revealing new mechanisms”.
“One of the biggest challenges in colorectal cancer is metastasis. This is the point at which most patients die. Something that tells us more about that and maybe indicates how this could be controlled is very promising,” he added.
Dr Sailem explained: “Cancer is not one disease - it can be classified into many diseases depending on tissue type and origin. We can take cells from diseased tissue and look at what the genes in these particular cells are doing. We can then identify genes to target for therapy – or genes for which targeted therapies already exist.”
THE LARGER TREND
AI and machine learning is increasingly being used to accelerate the development of targeted therapies in cancer and other diseases, with leading technology and pharmaceutical companies forming high profile partnerships in recent months.
One such collaboration between Novartis and Microsoft was announced in October to transform medicine with AI. Vas Narasimhan, CEO of Novartis, said, “As Novartis continues evolving into a focused medicines company powered by advanced therapy platforms and data science, alliances like this will help us deliver on our purpose to reimagine medicine to improve and extend lives. Pairing our deep knowledge of human biology and medicine with Microsoft’s leading expertise in AI could transform the way we discover and develop medicines for the world.”
ON THE RECORD
Professor Tim Maughan, professor of clinical oncology at the University of Oxford and advisor to Bowel Cancer UK, said Dr Sailem’s study linked to his own research into how cells within tumours ‘talk’ to each other.
He said: “What they say to each other is determined by molecular make up but also by the conversation going on between the cells. The shape that the cells have, the way that they are organised, the distance they are apart, how close the immune cells get into the cancer, is all a result of the conversation going on between the different cell types within a cancer”
“Dr Sailem in this study has found that in addition to identifying whole new genes which are important in bowel cancer, she has also picked up that genes that are part of that olfactory smell system play a part of this conversation.”
Commenting on the research, Professor Lawler said: “It is saying something about why there are olfactory genes in other parts of the body and how they might be responding to the microbiome in the gut. It will be interesting to see what stimulates these genes to up regulate or down regulate in their environment. From there we may be able to identify important biomarkers.”
He added: “Big data for better health makes sense. You can use that data to change lives, diagnose patients earlier, develop better treatment and improve their quality of life and above all, data can really save lives.”
For more information on bowel cancer go to www.bowelcanceruk.org.uk.