Knowledge integration is the key to variant interpretation


needle in a haystackOver the last few weeks I have discussed how new knowledge about our genetic make-up is fundamentally changing genetic medicine and genetic testing. We now know that rare variants are common and mostly harmless, and that we need to consider genetic variants innocent unless proven guilty.

But the key question in genetic testing today is: How do we recognise the genetic variants that do cause disease?

It’s a classic problem. Trying to find a needle in a haystack.

 

Keeping our eyes on the prize

We perform genetic testing to identify variants that are useful in the diagnosis, management or prevention of disease. We must keep our focus on this objective. On finding the variants of clinical relevance. We spend far too much time on variants with no clinical relevance. We often wallow at the bottom of the haystack, laboriously scrutinising stalks of hay to confirm they are hay.

 

Variant profiling would help interpretation

So how do we switch focus? To continue with last week’s crime analogy, offender profiling may serve as a useful template. Offender profiling is a method of identifying the most likely type of person that could have committed a crime based on evidence and information found at the crime scene along with specific characteristics of the crime itself.

Variant profiling would involve defining the type of variant that could cause a disease taking into account the characteristics of the gene, the disease and how variants in the gene can cause the disease. This can be defined ahead of testing, ideally with expert input and consensus. When a genetic test is performed, variants that match the pathogenic variant profile would be prioritised for evaluation, starting with variants that have the closest match.

So what should we include in variant profiling?

 

Diverse information can inform variant profiling

We could use a wealth of diverse information to build up detailed, discriminatory variant profiles for most disease genes. Gene-based, variant-based and disease-based information can all be useful. The basics will be applicable to many different genes, and much of the required data can be populated automatically. So we won’t have to reinvent the wheel for every gene. We could construct standard profiles that represent the typical scenarios for how gene variants cause disease. These basic profiles could then be customised for individual genes.

 

Knowledge integration is the key to variant interpretation

We need a variant prioritisation system that targets human expertise where it is needed and automates everything else

Informally many genetic testing laboratories sort-of already do variant profiling, at least in some situations. It is a common sense approach to bring together all the relevant information when you are trying to solve something. But the concepts and the data are being used variably, inconsistently, extemporaneously and sometimes incorrectly. Bringing formality, transparency and consistency to the approach would be very helpful.

Variant profiling is just giving a name to our urgent need to underpin variant interpretation with a transparently assembled, integrated knowledge base.

 

We need fast, high-capacity variant interpretation

Variant interpretation today remains an expert-driven, time-consuming, largely manual process that is performed separately by each centre that happens upon the variant. It is a major bottleneck in genetic medicine.

We are getting better at sharing the decisions we have made, through platforms such as ClinVar. And at describing the manual variant curation processes used. But it is still a reactive, labour-intensive, non-automated, non-scalable process. A process rooted in the old world of genetic testing.

We need to look beyond upscaling the interpretation processes developed in the era of small-scale testing of a few genes in a few people. We need to think, from the ground-up, about how we can marshal the wealth of available information into a variant prioritisation system that targets human expertise where it is needed and automates everything else.

Developing the intellectual concepts and the foundational frameworks that provide this will be essential to developing the fast, high-capacity, dynamic, scalable variant interpretation process genetic medicine needs.

Save

Save

Leave a Comment

Your email address will not be published. Required fields are marked *