Mapping bacteriophages in the gut

The human gut microbiome is a complex microbial environment that is closely linked to human health. Research has linked microbial imbalances in the gut to the development of diseases such as obesity, type 2 diabetes, Alzheimer’s and immune disorders.

Researchers have been studying the microbial environment in the gut for more than 10 years, and one recent advance is mapping the composition of the thousands of bacterial species.

Researchers were previously unable to determine whether the gut virome contains bacteriophages but have now finally solved this problem.

Using an advanced algorithm, researchers can identify the composition of viruses in a stool sample. This enables them to identify more effective antibiotics and to develop treatments for many gastrointestinal diseases.

“For example, we can identify which bacteriophages kill pathogenic bacteria. These may be used therapeutically to cure inflammatory bowel disease resulting from a bacterial infection or infections elsewhere. The right bacteriophages can also be used to combat antibiotic-resistant bacteria, which are currently difficult to eliminate,” explains a researcher behind the study, Simon Rasmussen, Associate Professor, Novo Nordisk Foundation Center for Protein Research, University of Copenhagen.

The research has been published in Nature Communications.

Bacteriophages were almost impossible to study

Bacteriophages attack bacteria in the same way that some types of viruses, such as SARS-CoV-2, attack people.

Although many people are not familiar with bacteriophages, researchers estimate that they are by far the most abundant type of cell on Earth. Bacteriophages are everywhere – in humans, in animals, on trees, in the soil and in water.

Nevertheless, bacteriophages are very difficult to cultivate in the laboratory because specific bacteria are needed to grow specific types of bacteriophages. Cultivating a specific bacteriophage requires knowing in advance which bacterium it infects.

Since there are tens of thousands of types, knowing where to start has been an almost unimaginable task.

“New methods for identifying bacteriophages in a sample were required, and we have now developed them,” says Simon Rasmussen.

Studying bacteriophages using algorithms and artificial intelligence

Simon Rasmussen and colleagues stepped out of the laboratory and instead analysed the vast existing metagenomics data on the composition of genes in stool samples from people, which researchers have used to determine the composition of bacteria.

The researchers used powerful computers and intelligent algorithms to examine all the genetic sequences for pieces of DNA that originate from bacteria. The researchers then assembled the pieces of DNA to construct whole bacterial genomes for characterising the individual bacteria.

“The bacteria in a stool sample cannot be identified by using a microscope. In the past 10 years, however, artificial intelligence and algorithms have been used to search for genes from bacteria in the samples. We have further developed these algorithms to be used on bacteriophages,” explains Simon Rasmussen.

Algorithms trained to recognise bacteriophages

Developing the new algorithms has not been easy.

Imagine wanting to create an algorithm to recognise cats. This can be achieved by showing it thousands of pictures of cats and teaching it to recognise the animal in a process called supervised learning.

But this technique cannot teach an algorithm to recognize bacteriophages because researchers do not have similar access to thousands of genomes from bacteriophages, which they can present to an algorithm.

Instead, the researchers had to use unsupervised learning, without explaining to the algorithm in advance which patterns in the bulk metagenomics data the algorithm should seek.

However, the researchers had access to more than 100,000 publicly available analyses of the microbial composition of human faecal samples. The algorithm trained on 650 of them, so that it can now determine very accurately the bacteriophages in a sample.

“The smart thing is that we can do this without purifying the virus in the samples. We just analyse the data as it is presented, and based on this, our algorithm can determine which bacteriophages are in the sample,” says Simon Rasmussen.

Bacteriophages as a precision weapon

Simon Rasmussen explains that their algorithm enables researchers to map the composition of bacteria and bacteriophages in the same sample and determine which bacteriophages attack which bacteria.

Since the bacteriophages are part of the chemical conflict between bacteria, they also determine the composition of bacteria in the human gut.

According to Simon Rasmussen, this conflict has therapeutic potential – the obvious type being antimicrobial resistance.

One disadvantage of using antibiotics is that they are very indiscriminate microbiological weapons, destroying everything.

Bacteriophages can fight specific types of bacteria and usually only attack one species, so they can be precision weapons.

“Our algorithm enables us to determine which bacteriophages we need to combat a specific infection with very specific bacteria, such as in the gut,” explains Simon Rasmussen.

Can be used therapeutically

Insight into microbiological warfare may also be used to adjust the composition of gut bacteria.

Some bacteria are associated with the development of various diseases – sometimes directly caused by a specific bacterium, such as in inflammatory bowel disease – but they can also be associated with negative effects on health just by being present.

People would like to eliminate these bacteria and replace them with health-promoting bacteria.

“Much research focuses on how faecal transplantation can change the composition of gut bacteria and cure various diseases. The latest research shows that transplanting whole faeces is not necessary. Transplanting the bacteriophages can thus be used to manage the composition of gut bacteria to end up with the bacteria that promote health,” concludes Simon Rasmussen.