The gut is home to billions of bacteria, known as the microbiome, however, little is known about how this microbiome changes over time. An international team of researchers, led by InSilico Medicine, a US-based artificial intelligence startup has teamed up to figure out exactly that. In a new paper, not yet formally published but available as a preprint on bioRxiv, the authors suggest the human microbiome could be used as an amazingly accurate biological clock.
The gut microbiome has become increasingly recognised as an important indicator of health and disease, and links between specific microbes and host health, genotype, and diet have been identified. Furthermore, this complex ecosystem is responsible for performing multiple important functions involving the digestive system and immune system.
The researchers used machine learning to analyse more than 3600 gut bacteria samples from 1165 healthy individuals around the world obtained from 10 publicly available datasets. The data were then divided into three age groups: 20 to 39 (young), 40 to 59 (middle-aged), and 60 to 90 (old). The deep learning algorithm based on regression was trained using 95 different species of bacteria ― defined as important intestinal biomarkers of human ageing based on so-called Permutation Feature Importance ― found in 90 per cent of the samples (the training dataset) to determine their correlation with age.
The algorithm was able to accurately predict the ages of remaining samples (10 per cent) to within four years. To verify the method, the results were also compared those obtained using three traditional classification methods ― random forest, support vector machine, and elastic net regressor ― which were all found to perform poorly in comparison (errors exceeding 11 years).
According to the authors, microbiological profiles can be used to predict human ageing. The findings showed that 39 out of the 95 types of bacteria were most important for predicting age. Furthermore, certain bacteria were found to increase with age, for example, Eubacterium hallii, which is often associated with intestinal metabolism, whereas other bacteria, such as Bacteroides vulgatus ― one of the most common microorganisms in the human gastrointestinal tract ― was shown to decrease.
One challenge the researchers are aware of is the diversity of microbiomes around the globe, therefore, to make the tool more accurate, studies will need to be performed using data from different populations. Dr Alex Zhavoronkov, CEO of Insilico Medicine and who led the study, told Science magazine the scientists are unsure whether changes in the microbiome lead to ageing or if these changes are just one of the side effects of ageing, but says “Age is such an important parameter in all kinds of diseases.”
The new “ageing clock” could be combined with other biomarker predictors, such as telomere length ― structures at the end of chromosomes that become shorter with age ― to paint a more precise picture of an individual’s health or biological age. The new model could also be used to investigate how factors like alcohol, antibiotics, probiotics, and diet affect longevity, and whether certain diseases alter the microbiome.
(1) Fedor Galkin et al. Human microbiome aging clocks based on deep learning and tandem of permutation feature importance and accumulated local effects. bioRxiv (2018) DOI: 10.1101/507780