For many years, scientists have focused on how the diversity and composition of gut bacteria affect health, revealing that some bacteria are associated with health and others are linked to disease. A new study, however, highlights that the number of microbial cells per gram – the microbial load – also has a crucial role.
The fact that gut bacteria are integral to overall health has become increasingly evident over the past decade. The composition of gut bacteria has been associated with the risk of developing type 2 diabetes, obesity, cardiovascular diseases and even neurodevelopmental disorders such as Alzheimer’s disease and Parkinson’s disease.
Although much research has focused on how types of gut bacteria affect health, the new study emphasises that microbial load is likely of equal, if not greater, importance. The researchers also developed an artificial intelligence tool that determines the microbial load based on variation.
The study’s findings underscore the important role of microbial load in health, suggesting that previous research has underestimated this.
“We have traditionally focused in this field on the relationships between bacterial species, often neglecting the critical role of microbial load. For instance, we could determine that bacteria X comprises, say, 2% of the total bacteria in a stool sample, but we did not know whether this equates to 100,000 or 1 million individual bacteria. Our artificial intelligence tool addresses this gap by adding a crucial layer to the research – quantifying microbial load alongside studying the composition of gut bacteria,” explains a researcher behind the study, Michael Kuhn, Research Staff Scientist, Molecular Systems Biology Unit, European Molecular Biology Laboratory, Heidelberg, Germany.
The research, which was undertaken in collaboration with the University of Southern Denmark and the University of Copenhagen, has been published in Cell.
The first author is Suguru Nishijima, who was a postdoctoral fellow at EMBL at the time and is now an Associate Professor at the University of Tokyo, Japan.
Counting bacteria is difficult
The researchers behind the new study, like many others, have long focused on how the composition of bacteria in the gut affects people’s overall health.
They have participated in major international research projects investigating how gut bacterial composition affects the development of cardiovascular and liver diseases.
For these studies, the researchers used advanced sequencing techniques capable of determining the relative composition of bacteria. However, determining the microbial load required an additional step to count all the bacteria in a stool sample with labour-intensive experimental techniques.
This process is not only difficult but also time-consuming and prohibitively expensive.
“We therefore also sought to explore the potential of developing a machine-learning method to estimate microbial load using existing sequencing data,” says Michael Kuhn.
Artificial intelligence model trained on faecal samples
The researchers developed an artificial intelligence model to identify connections between the relative composition of gut bacteria and the microbial load.
To train their algorithm, the team used a data set comprising faecal samples from 3,700 individuals.
For these samples, both sequencing and microbial load data were acquired, enabling the artificial intelligence algorithm to align the bacterial composition and microbial load.
The model’s effectiveness was validated on a separate data set not previously encountered, confirming its accuracy and reliability.
“This breakthrough means that we no longer need to manually count bacteria in faecal samples to determine the microbial load. Existing sequencing data can now be used to derive both the relative bacterial composition and the microbial load. This simplifies and enhances research on how gut bacteria affect human health,” notes Michael Kuhn.
Identifying associations between microbial load and disease
After training their model, the researchers applied it to an extensive data set comprising stool samples from 27,000 individuals gathered from 159 previous studies conducted in 45 countries.
This vast data set revealed numerous factors that are associated with microbial load. Some were expected, such as diarrhoea and antibiotic use being linked to lower microbial load, whereas constipation is associated with higher microbial load.
Others were more surprising, such as women exhibiting a higher average microbial load in their stool than men, and younger people tended to have a lower microbial load than older adults.
The researchers also discovered that changes in microbial load were associated with various diseases.
Intriguingly, some diseases shared similar relative profiles in microbial composition, but these similarities were driven by parallel changes in microbial load across the conditions.
Further, the presence of certain bacteria in specific diseases was often more strongly associated with the microbial load than with the disease itself.
“This implies that changes in microbial load related to certain diseases may be more important in shaping the composition of gut bacteria than the diseases themselves,” says Michael Kuhn.
Model freely available but needs more training data
The researchers have made the algorithm they developed to study microbial load freely available, enabling all researchers to add an extra dimension to their work.
At the moment, this tool is limited to researchers studying gut bacteria. Although the method should be applicable to bacteria present elsewhere in the human body or even in external settings such as water or soil, where monitoring both bacterial composition and microbial load is likely just as crucial, training data are currently lacking.
Michael Kuhn emphasises that applying the model in various contexts requires training it on data specific to the relevant environment. Meanwhile, he remains focused on gut bacteria in his research.
“Until now, we have lacked important context in our studies. With this model, we can incorporate that context and more deeply understand the role of gut bacterial composition and microbial load. A fascinating next step would be to investigate what factors influence microbial load over time and how this affects health,” he concludes.
