Bioinformatics tool predicts cell transformation

Researchers at The University of Hong Kong (HKU) have developed a new bioinformatics tool for the precise prediction of cell transformation and disease formation, including cancers.

Over time, cells in the human body constantly change to different forms. Yet even though they have been studied for more than a century, it remains a challenge to understand what they do and how they change over time during normal health or ageing, or how these changes in cellular structure result in diseases including cancer.

Omics data analysis, which seeks to capture the rich information of single cells at high precision, is important for tracking cells transformation and hence for the prediction of occurrence and recurrence of diseases, and for survival analysis and discovery of biomarkers. An interdisciplinary team at HKU has now created a computational tool for studying omics data that overcomes current challenges in the study of normal and malignant cell transformation, which they have described in the journal Nature Communications.

“You can imagine cells are the mountain hikers travelling in a huge ‘landscape’,” said research leader Professor Kevin Tsia, Director of the Biomedical Engineering Programme at HKU. “The fact that they evolve into different cell states and types in normal health and in disease can be analogous to that the hikers travel along different trails (or trajectories) in this landscape. Our ultimate goal is to find ways to locate cells on this landscape, predict their trajectories and thus understand how our body develops normally and how it fails in diseases.

“Accurate prediction of these trajectories relies on the ability to measure and analyse all the relevant signatures representing each cell. This is like performing in-depth interrogation with each hiker so that we can gather his/her identity, accessories, physical/mental conditions and so on.”

A wide range of so-called ‘omics technologies’ are now available to capture rich information of single cells at high precision, such as a catalogue of genes that are turned on or off in cells or the protein types present in cells.

“However, current algorithms often cannot handle the sheer size of complex single-cell omics data (more than millions of cells),” said Dr Joshua Ho, a, co-author on the study. “They are also often designed to work well only for analysing only one specific type of cell signatures. Thus, we still lack a solution that is versatile enough to perform comprehensive trajectory inference.”

Overcoming these barriers, the new trajectory inference algorithm developed by the HKU team, called VIA, can process omics data up to 100 times faster than existing technologies, making it powerful in handling large-scale single-cell data (more than millions of cells). It can hence discover elusive cell lineages and rare cell fates in a variety of biological processes that can hardly be discovered by other methods, shedding important light into how disease evolves.

“Our method uses a novel unsupervised machine learning approach, which allows us to efficiently identify the best trajectories that represent the underlying biological progressions, no matter how complex the trajectories are,” said Shobana Stassen, first author on the study. “And we showed that VIA works very well, even when the cell population size is beyond what the current methods can handle.”

To ensure that VIA can be widely adaptable, the team carried out comprehensive investigations of different important biological processes, based on a wide range of data types, including single-cell proteomic, transcriptomic epigenomic and multi-omics datasets.

For example, VIA can reveal complex and subtle transitions of cell state during haematopoiesis — a process of blood formation from bone marrow. It also robustly uncovered the fascinating and intricate process at single-cell precision that turns a single fertilised egg into a whole new individual with all the organs (a process called organogenesis), even when the cell count to be analysed is beyond one million cells. This is the scale otherwise not affordable in many existing methods.

Another key outcome of this work is that the team explored the use of VIA to analyse the high-resolution cell image data that can infer physical traits of the cells — a cell signature that has largely been underexplored for trajectory inference. Apart from the omics data, cell image data contains rich information on how different cell types look and how the cell changes its outlook (morphology) over time.

Using an optical microscope system developed in Prof Tsia’s research lab, the team generated a large amount of single-cell image data at the speed at least 100 times faster than typical microscopes. They then used VIA to successfully reveal subtle and variations of cell mass and the mass content distribution over the cell cycle progression — a fundamental cellular process that maintains life, and might lead to cancer if the process goes wrong.

Meanwhile, research teams in the US and mainland China have adopted the VIA and the related computation tools developed by the team for COVID-19 research to track and predict immune responses after infections or vaccinations, and the body’s response to treatments.

“We are now applying VIA in different biomedical and biotechnology applications, in collaboration with researchers here in Hong Kong as well as overseas,” Prof Tsia said. “Tip-of-the-iceberg examples include tracking anticancer treatment efficacy, as well as accelerating the drug discovery process in pharmaceutical industry.”

“VIA is an open-source tool and we made it publicly available,” Stassen added. “We are constantly improving and upgrading the tool. I hope VIA could widely benefit the community of biologists and biomedical scientists who investigate different aspects of biological evolution.”

Image caption: Dr Joshua Ho and Professor Kevin Tsia demonstrate their bioinformatics tool.

Please follow us and share on Twitter and Facebook. You can also subscribe for FREE to our weekly newsletters and bimonthly magazine.