In 2019, over 1,000 volunteers searched for viruses inside infected cells; in 2 months, they made 90,000 classifications

Finding and identifying viruses inside infected cells is far more complicated than spotting them under a microscope. Although modern imaging techniques can create and deliver super-detailed three-dimensional images of infected cells, it is so full of tiny details that it takes weeks for researchers to isolate every virus.

So as a solution, in 2019, scientists decided to try something else. Rather than leaving the process entirely in the hands of experts and professionals, they invited the public to help through an online citizen-science project called Science Scribbler: Virus Factory. Here, people were asked to look through the images of the infected cells and indicate the locations where they thought the virus particles existed. The responses exceeded expectations. In just 2 months, over 1,000 people participated in the project and made over 90,000 classifications for the researchers to start their analysis.

Turning a complex scientific task into a citizen-science project


The project could only analyze 3D images made using cryo-electron tomography, the technique that makes it possible for the researchers to see frozen specimens in three dimensions with high definition. Indeed, the images yielded necessary data about the structures called virus factories, specific parts of the cell where the virus is produced.

But the challenge was not just to collect pictures but to interpret them. Each 3D volume contained numerous virus particles that were tiny, densely packed, and often looked similar to each other, and identifying and locating each of them would require a great deal of human effort and careful attention. Instead of asking the volunteers to do such complex scientific measurements, scientists came up with something easy. They displayed small bits of the images to the volunteers through the internet and asked them to click the points in which they found viruses. Because numerous people examined the same images, scientists could compare their answers and combine repeated observations to improve accuracy.

The project quickly attracted attention for participation. In just two months, over 1,000 volunteers (people, not the scientists) had produced more than 90,000 classifications. Each individual’s contribution was a classification for a very small portion of the whole dataset, but together they gave a detailed and extensive map of the locations where viruses were found in the infected cells. Scientists grouped many clicks made by the volunteers, filtered the data, and produced an initial classification of all viruses found in the three-dimensional structure. This information was quite helpful for researchers because it would take them much more time to get the same results via manual annotation. The project also proved how large groups of volunteers can collectively solve scientific problems that are difficult to automate completely.

Why does the data matter

The volunteers were not simply counting viruses. Their work helped scientists understand the virus maturation within infected cells and the ways viruses transform during their maturation process.

The spatial relationship between the locations of viruses is an important step towards understanding how they utilize the machinery of the host cell in order to multiply. Scientists can study different stages of virus maturation present in certain locations in the factory of viruses and how the infection proceeds. The data collected by the volunteers was also used to train computer algorithms. The crowd-generated location insights and classifications were used to create deep-learning systems that can recognize, separate, and categorize viruses in difficult biological photos. This showed how citizen science would work together with AI to speed up the analysis of a vast amount of biological imaging information without requiring much from specialists.

The power of small contributions by volunteers

Science projects, like Science Scribbler: Virus Factory, show that sophisticated laboratory equipment and public participation complemented one another. If tasks are carried out in the right manner as the volunteers helped where automated analysis remained difficult.

The success of the 2019 project shows that the power of small contributions brings impressive results. More than 1,000 volunteers, each making simple observations, collectively generated over 90,000 classifications that helped researchers map viruses inside infected cells.

Those data not only advanced the study of virus factories but also supported the development of machine-learning tools that can accelerate future biological research. Today, with the growing number of data generated by new imaging technologies, such cooperation between scientists, commoners, and artificial intelligence gains importance.