Moore’s Law and Crystal Balls

When developing analytical methods to quantify bioactive trace components in foods, I’ve always tried to work at the edge of current methodology. But for some time now, a big question has loomed over me: when can I expect to overcome the limitations of these methods? I set out to provide an answer by predicting when the whole food metabolome will be identifiable and detectable...

First, we separated the unknown components of the food metabolome into three different sets of “dark matter:”

Set 1. Unknown metabolites not yet present in databases
Set 2. The metabolites not yet detectable with current equipment
Set 3. Molecules with unknown structures.

We then looked at the most important compound databases from different organizations. The number of compounds included in these databases ranges from a few thousands (Golm Metabolome Database) to almost a hundred million (PubChem), running the full gamut from primary metabolites in humans to all man-made chemicals. In principle, all these compounds could be present in foods, but the majority of them are very unlikely to appear.

In fact, for our purposes, the databases of PubChem, ChemSpider and Metlin should not be included as they mainly contain xenobiotica. A much better estimation comes from the ~20,000 compounds in KEGG or the currently updated human metabolome database (HMDB), which contains around 114,000 compounds.

Considering that MS sensitivity and resolution appears to be increasingly exponentially, we leaned on laws governing advances in general technological developments to make our predictions; namely, the laws of Moore (the number of transistors in an integrated circuit doubles every two years) and Kurzweil (the law of accelerating returns, estimating that the intelligence of artificial intelligence will surpass humans in around 2045 – an event he rather wonderfully refers to as the “technological singularity”).

Our predictions? The unraveling of dark matter set 1 by 2025, set 2 by 2032, and set 3 by 2041.

The important question for dark matter set 2 in my eyes: when will these metabolites be detectable by analytical equipment as resolved features? To get there, we’ll need an expansion of non-targeted metabolomics into single molecule detection. With the exponential development of current limits of detection (LOD) of 6 x10^-18 mol to single molecule detection (1.66x10^-24 mol) – we believe that single molecule detection on routine MS equipment should be feasible around 2032.

Such advances will allow us to uncover and assess what I would call the “iceberg” of all food metabolites (we see but the tip at present). As we move through our predicted timescale, researchers will be increasingly able to map the bioactive compounds – both toxic and health-promoting.

As we approach 2045, our coverage will be comprehensive – with a concomitant high impact in food safety and regulation. And so, in a way, our work could be seen as a valuable tool for proactive regulators and risk assessors...

The advancement of any other analytical field could be predicted in a similar way – perhaps even more easily; after all, food matrices are some of the most complex around. I would encourage anyone that is capable of such extrapolation to give it a shot – the information gleaned could guide researchers forward with a clear view of the technical capability milestones we should expect to reach.

If you’re interested, you can watch my talk on this topic at RAFA 2020 here.