Biomarker discovery: avoiding the oversimplification trap

Why the search for the silver bullet does not lead to adequate results in biomarker research

In science, as in life, we like to find the simplest path to the answers we seek. We look for the most efficient route from point A, our hypothesis, to point B, our proof.

Euler’s famous K?nigsberg bridge problem is a great demonstration of the power of simplification: he showed that if you want to get from one K?nigsberg island to the other by crossing all seven connecting bridges once, you’ll find that the task is impossible even - and actually more simply - if you remove all other features from the map and look only at the bridges.

When we apply this lesson to scientific research, we aim to simplify our approach by focusing on certain hypotheses and eliminating approaches that could distract us from the essential. Reducing complexity can often help us see things more clearly.

However, simplification can also lead you down the wrong path. If oversimplification results in the wrong conclusions in biomarker research, it means that the biomarker – and potentially the therapy that it would stratify – is destined for failure.

I like to think about therapies as taking patients from point A, where they’re experiencing symptoms, to point B, where they’re hopefully doing better. If you jump in your car to get from a point A to a point B, you’d never head to your destination in a straight line. You have to follow the roads, stop at traffic lights, and watch out for pedestrians or speed bumps that might get in your way. The actual length of your drive may be several times further than the air-line distance. Look at the picture above: if you want to travel from your office at place A and have a picnic with friends at place B, you’d have to wiggle your way around the available paths. You would never consider just walking from A to B in a straight line, because you know it would get you into all sorts of trouble.

If we accept this as an obvious truth when we navigate in a physical sense, why should we treat patients like there were no obstacles on the road to therapy success? A quote I read recently sums it up well: “You can’t cheat to cure a disease.” (Potential fabrication in research images threatens key theory of Alzheimer’s disease | Science | AAAS). The article in which it appeared was about scientific misconduct. While I’m not suggesting that researchers who seek the shortest route are committing misconduct, there is no way around the fact that oversimplification can be a threat to good science. Occam’s razor does not always apply – the simplest explanation just is not always correct. We must acknowledge biology as being complex to the point that the simplest answer is usually not going to be correct.

What’s the way to deal with the complexity in biomarker discovery, then?

This doesn’t mean we need to throw our hands up and accept defeat. I am also not suggesting you should do any and every test you can imagine, hoping to land on promising results. However, you clearly need to do more than the reductionist approach of only testing biomarkers that are directly related to a drug target or well established in the respective fields. If you consider the whole body of scientific work associated with your topic of interest, you will almost inevitably have to conclude that a silver bullet biomarker probably doesn't exist - and that body of work should guide your approach to biomarker research.

Modern omics technologies provide a great opportunity to gain a deeper understanding of biological systems and the action of interventions therein. Some have described these technologies as potentially providing a “GPS system” that leads you toward your goal no matter where you start, and no matter what obstacles you face along the way.

Of course, it’s never quite that simple. Biology is still complex, and omics technologies vary considerably regarding type of analytes, coverage, technological features, data formats and the like. Nonetheless, omics technologies embrace the complexity of biology rather than ignore or negate it. Together with powerful computational methods, this provides the basis for truly personalized medicine.

Perhaps you’re struggling to imagine that the best biomarker panel to stratify patients for a particular therapy is not along the direct path between A and B. Don’t make the same mistake as Albert Einstein. Despite defining the formula that provided the theoretical basis for the existence of black holes with the theory of general relativity, he did not believe they existed in the real world. Einstein was limited by his own imagination. Everyone makes mistakes, but you should avoid the mistake of discounting the potential of omics.

Which omics technology do you think has the greatest potential? What are success factors to successfully identify and translate -omics based biomarker signatures? Have you seen examples where the less simple route has yielded more interesting results? Let me know what you think in the comments.