Q: The Phenograph is an important underpinning of BioSymetrics’ machine learning-enabled, data-driven approach to advancing precision medicines. What is it, exactly?

Gabe: The Phenograph is a linkage map that allows us to connect phenotypes to genes to diseases to drugs, in any direction, and across multiple model systems, allowing us to quickly translate findings between model systems and humans.

Q: What is the value of a linkage map like the Phenograph?

Gabe: Essentially, it allows us to go more quickly from what we see in the clinic to what we see in the lab. It happens all too often, in clinical work, that you’ve identified some clinical phenotype that you think is relevant to a certain patient population, but then it’s hard to take the next step of translating that data into something that can impact patient care. That barrier to translation, bringing what you see in the clinical population to what you can model in the lab, stops a lot of discovery programs.

The Phenograph allows us to traverse more readily among model systems, forwards and backwards, in any direction we choose or starting with whatever asset we have (whether it’s starting with a phenotype or gene or drug). We can, ultimately, be much more efficient about how we translate research discoveries, enabling quick transitions from results in the clinical space into a model system, and vice versa.

Q: That is powerful. What kind of clinical data and model systems does the Phenograph contain today, that we can draw links between and better progress discoveries as a result?

Gabe: We link human clinical phenotypes, from human phenotype ontology or ICD codes, to human genes and disease labels. This is a big part of what our data science teams focus on, curating and creating these associations. On the model systems side, in zebrafish, for example, we have all kinds of experimental datasets, including those we’ve generated ourselves, that inform relationships on gene function.

Q: We’ve talked a lot about phenotypes. Can you give an example of how we investigate a phenotype using the Phenograph?

Gabe: Sure. Right now, one of our data scientists is doing this for heart failure. They start with the “heart failure” disease label and, within that, associate several phenotypic terms to it. Some examples of phenotypic terms associated with heart failure might be “fatigue” or “reduced ejection fraction” which is something you would see in an echocardiogram. Basically, any observable, physical characteristics associated with “heart failure.” Our data scientist then maps that to human genes and the corresponding zebrafish phenotypes, our chosen model system, and describe them as “normal” and “abnormal.” We can then use that zebrafish to experimentally confirm new gene-disease associations in vivo, and test out the effects of new compounds. The goal is to move quickly from human phenotype to model system phenotype to potential drug, and then ultimately back to the clinic!

Q: You answered all the obvious questions. What else would you add, in describing the Phenograph?

Gabe: I would add that the key to the Phenograph is the underlying linking of the phenotypes. My mantra for a long time has been that the phenotype, or observable physical characteristic of a disease, is all that matters. In science, we seek to understand genes so we can improve phenotypes. No patient cares if they have a disease label or a disease-causing gene – they want to improve how that disease is expressed, as in, their phenotype.

A phenotypic-centric approach brings us much closer to helping patients by addressing their lived experience first, and we can be more efficient in this way to improve human health.