Big data is often a characteristic of microbiome studies. Still, our predictive power is poor, according to Hector Garcia Martin from the Joint Bioenergy Institute. His research focuses on taking all that “omics” data and distiling it down to information we can apply. The traditional equations for studying enzyme kinetics don’t work particularly well when you’re looking at thousands of proteins at once in a real-world environment, so instead, Hector uses a combination of proteomics, metabolomics, and machine learning.
The problem is that machine learning requires a LOT of data – both many strains of the target microbe and many replicate experiments. Luckily, two tools have emerged to solve these issues in recent years – CRISPR and robots. CRISPR-Cas9 genetic modification quickly creates many bacterial strains, while microfluidics allows more experiments to run concurrently than a human could set up. This approach will also work for species abundance, which would be applicable to microbiome dynamics.
The best news is that Hector is looking for collaborators as they seek to continue and improve these techniques! We don’t need machine learning to predict that we’ll be seeing more from this research soon.