Healthcare

AI helps explain your microbiome

Share this post:

Can the tiny microbes on your leg reveal your age, or that you smoke? Or that you are menopausal?

Billions of microbes live on our skin, they help maintain skin condition, they are our first line of defense from external pathogens and can impact how we respond to treatment.

These microbes, which are one tenth the size of human cells, are part of the human microbiome which consists of the collective genome of microbes inhabiting the human body, including bacteria, archaea, viruses, and fungi. A better understanding of our microbiome could help improve overall health and wellbeing and accelerate the development of personalized treatments (including prebiotics, probiotics, and postbiotics).

 

Microbiome

Model explanations for the 62 first samples taken from the Canada cohort.

This was the motivation behind a new publication appearing today in the peer-reviewed journal Scientific Reports. Scientists from IBM Research, the Science and Technology Facilities Council’s Hartree Centre, and Unilever report a new Explainable AI (EAI) framework that unlocks the “black box” of AI to help enable explainable data analysis and interpretation. This has now been applied to understanding the link between skin microbiome composition and personal wellbeing.

In the first ever such analysis of the cosmetic skin microbiome, the team analyzed over 1,200 skin samples, from more than 160 women in North America and the UK, using the EAI framework to identify differences in microbial composition linked to skin dryness, menopausal status, age and smoking habit.

Using the EAI framework the researchers were able to predict skin dryness with 90 percent accuracy, whether a woman is pre- or post-menopausal with 92 percent accuracy, smoking habit with 85 percent accuracy and age with 85 percent accuracy, all just from the composition of their skin microbiome, offering new insights that can potentially accelerate the development of personalised treatments for healthy skin.

Our findings on the predictive power of the skin microbiome point to the possibility of using readily accessible microbiome samples to investigate a broad swathe of human wellbeing endpoints.

One of the biggest challenges associated with microbiome analysis includes noisy, sparse and high dimensional datasets, with many more microbial features than samples. As such, the extraction of a few important microbes, among thousands, that are mostly associated to a certain condition or characteristic of the host organism (e.g, human body) is a very difficult computational task.

In addition to data, it was important to develop the EAI framework with transparency, meaning not only does it predict characteristics (e.g., menopausal status) from microbiome data, but it also explains how it came to the predictions. The explanations are expressed in terms of changes in the abundance of key microbes (features) associated with a specific characteristic status (e.g., pre-menopause or post-menopause).

Our explainable AI framework has introduced a new approach to analyse microbiome data and infer actionable insights, which can be used to augment standard bioinformatics approaches. While the framework was applied to skin microbiome data in this example, it could be applied to help predict a condition or trait from microbiome data of the person, which has applications in personalised healthcare.

We hope the scientific community expands and builds on this initial work as the explanations may offer new insights into the complex interactions between microbes and human wellbeing, possibly leading to associations with new uncharted benefit areas and well as providing new intervention targets and strategies to modulate the microbiome for improved health outcomes.

This research was part of the Science and Technology Facilities Council’s Hartree Centre’s Innovation Return on Research programme.


Anna Paola Carrieri, Niina Haiminen, Sean Maudsley-Barton, et al. Explainable AI reveals changes in skin microbiome composition linked tophenotypic differences. Sci Rep (2021). https://doi.org/10.1038/s41598-021-83922-6

Inventing What’s Next.

Stay up to date with the latest announcements, research, and events from IBM Research through our newsletter.

 

Research Staff Member, ML & AI for Life Science and Healthcare, IBM Research

Michael Hoptroff

Senior R&D Manager at Unilever

Barry Murphy

Microbiomics Science Lead at Unilever

More Healthcare stories

Pushing the boundaries of human-AI interaction at IUI 2021

At the 2021 virtual edition of the ACM International Conference on Intelligent User Interfaces (IUI), researchers at IBM will present five full papers, two workshop papers, and two demos.

Continue reading

Deep Document Understanding: IBM’s AI extracts data from complex documents

Using novel deep learning architectures, we have developed an AI that could help organizations, enterprises, and data scientists to easily extract data from vast collections of documents. Our technology allows users to quickly customize high-quality extraction models. It transforms the documents, making it possible to use the text they contain for other downstream processes such as building a knowledge graph out of the extracted content.

Continue reading

AI Hardware Composer launches on two-year anniversary of IBM AI Hardware Center

Unveiled at the two-year anniversary of the IBM Research AI Hardware Center, AI Hardware Composer for analog AI hardware enables one to master and accelerate the AI hardware technology to power more sustainable AI models. It’s one of many upcoming developments of the AI Hardware Center, launched in 2019 to innovate across materials, devices, architecture and algorithms.

Continue reading