Genetics: “We try to simplify diagnosis for rare diseases“

Interview with Dr. Peter N. Robinson, Institute for Medical Genetics, Charité University Hospital Berlin

Sometimes your TV is actually right and diagnosing an illness is really a puzzle. This is the case with rare diseases for example, which only affect a small portion of the population. Physicians are then confronted with the problem of not having enough experience with a specific illness and its symptoms to be able to make a diagnosis.


Photo: Gold piece surrounded by black ones

Orphan diseases affect only very few people; © Priewe

At least in the real world, physicians cannot count on luck or coincidence to come to their rescue. A new diagnostic procedure that reconciles data from genetic analysis with disease symptoms and suggests possible diagnoses could assist them in the future. spoke with Dr. Peter N. Robinson about a procedure called “PhenIX“ and its application.

Dr. Robinson, what does the name “PhenIX“ stand for and what is behind it?

Peter N. Robinson: “PhenIX“ stands for “Phenotypic Interpretation of eXomes“. We are trying to simplify diagnosis for so-called rare diseases. In genetics, pediatrics and several other disciplines we come across people with these rare diseases. As of today, we know of approximately 7,000 of them, but there are probably 1,000 other diseases that are still not known at this point. The diagnosing process is difficult, because there is no physician that can know all of these diseases and their symptoms. Diagnosing a disease often works by pattern recognition of symptoms, which represent the phenotypic characteristics of the illness. Coughing, a runny nose and fever suggest a common cold for example. This way of diagnosing rarely works with rare diseases, because a physician would probably have rarely seen these patterns during his or her career.

How can PhenIX help in this case?

Robinson: In genetics, you compare the phenotypic patterns with databases and scientific literature. You search for diseases that have a similar pattern. There are a number of databases, like the Human Phenotype Ontology (HPO) for instance, which was developed by my task force in 2008. This semantic network specifies the different phenotypes and their relationships between each other. HPO is being used by PhenIX to analyze the clinical symptoms.

The second step is the clinical exome, where we include all of the currently known Mendelian genes that cause rare diseases. So far, those are about 3,000 of the approximately 20,000 known human genes. We have determined them through so-called next generation sequencing. This highly effective genetic analysis method has been around for nearly a decade. By now, this process is used in routine diagnostics in the U.S., the United Kingdom and the Netherlands, but not in Germany yet, because the Association of Statutory Health Insurance Physicians has forbidden its use. Yet it truly marks a quantum leap for the affected patients.
Photo: Puzzle in pieces, a stethoscope lying on top

Looking for the right diagnosis can sometimes be like solving a puzzle for the physicians involved. "PhenIX" helps them to put the pieces together; © Benoit

After all, PhenIX filters out approximately 100 genes from a patient’s genetic makeup that are possible candidates for causing a disease. Based on the patient’s phenotype and his or her symptoms, we are able to prioritize these candidate genes. Ideally, the correct disease gene subsequently ends up on the top of the list.

Where does the data for the HPO come from?

Robinson: My group has worked on this project since 2007. We are supported in this by international groups: these include Online Mendelian Inheritance in Man (OMIM) of the John Hopkins University and the Undiagnosed Diseases Program of the National Institutes for Health, both from the U.S., Orphanet, a European database for ”orphan diseases“, meaning rare diseases, in Paris, and a number of clinics in the U.S. and Europe. These colleagues and my own team have compiled data on diseases and phenotypes over the past seven years. We have updated the HPO more than 3,000 times since 2007; so it truly is a very active project.

How reliable is the diagnosis?

Robinson: I differentiate between an expert system and a system for experts here. IBM Watson would be an expert system for example: it works like a robot; you input something and the system gives you an answer.

A system for experts is only meant to help experts by providing the required information and presenting it correctly. Our system is such a system for experts. Without skilled physicians, it would be useless: you have to be able to input terms from the HPO and determine a phenotype. Your subsequent output is a list with ordered differential diagnoses. We were able to show that the correct diagnosis is very frequently ranked on top of the list when using PhenIX. Previous systems would have only provided an unordered list. Although the correct gene would be included in this list, it could take days to find it.
Photo: Cotton swap and speciman cup on a print out

Using a specimen from the patient, "PhenIX" makes suggestions about genes that possibly cause the disease; © Simon

Using PhenIX, we have also examined patients here at the Charité, who have been tested without avail sometimes for years using classical Sanger sequencing, chromosome analysis and other methods. We were successful in providing a diagnosis for many patients, who previously lacked one.

Nevertheless, PhenIX is a tool that a physician can utilize as a part of diagnostics. If he or she receives a diagnosis using this method, they still have to validate it with the usual methods and deliberations. That is to say, he or she specifically examines the patient’s entire family and researches literature on the suggestions given by PhenIX.

Can this method be used by other hospital facilities and if so, how?

Robinson: Next generation sequencing is essentially feasible wherever there is a respective gene-sequencing device. There is international consensus that this technology is significantly better and more cost-effective than conventional diagnostic methods. Nevertheless, the Association of Statutory Health Insurance Physicians has forbidden its use for routine diagnostics in Germany, which is something I just do not understand. This is why it is primarily only research institutes that use this method in Germany, yet it could be these institutes that could conduct the analysis as well. We provide software on the Internet for this free of charge. Once the German billing problem has been solved however, any general practitioner could send us or another center a sample, so that we could launch a diagnostic investigation.
Photo: Timo Roth; Copyright: B. Frommann

© B. Frommann

The interview was conducted by Timo Roth and translated from German by Elena O'Meara.