You are here: Home / Supporting the NHS Genomic Medicine Service / Finding a genomic cause: the clinician’s role

Finding a genomic cause: the clinician's role

If you’re a clinician requesting genomic testing for a patient, there are some important things for you to know and do in order to maximise the chance that a diagnosis can be found by colleagues analysing the results.

Once a sample has been collected from a patient and genomic sequencing has been performed, scientists are faced with a sea of the individual’s genomic data.

To try and find the genomic variant responsible for disease, they use computer-programming methods and other information to whittle down the genomic data in stages: a process known as variant filtering.

These stages fall into two categories: those that are applied automatically – those that you cannot influence – and those that are based on information you provide: those that you can influence.

Where you come in

Early filtering stages are based on information about the human genome and health at a population level – so-called ‘big data’ – and are out of your control.

For the latter stages, however, the information and samples you provide can make a huge difference to the outcome.

How is the data from genomic sequencing filtered down?

 

Trying to find the genomic variant responsible for disease is like trying to find the one, correct ball in a huge ball pit. Use the arrows above and to the right to scroll through the images and learn more about the process.

First, bioinformaticians in the laboratory begin by removing any variants that are not in the ‘coding’ region of the genome (shown as white balls)…

Non-coding variants, as they are known, are removed because, in order to try and find a diagnosis, we are searching for variants that we already understand to have an effect on proteins.

Next, all the ‘common variants’ (the turquoise balls) are also filtered out – since they are unlikely to be the cause of a rare condition, for which genomic sequencing is most often used…

As you can see, these first two automatically-applied stages are the most powerful in terms of volume and take us from around three million variants to a more manageable pool of approximately 250-300 that are classified as rare.

Next, tailored filtering stages are applied based on the information that has been provided by the referring clinician about the individual patient and, in some cases, their family members.

If family information is available, laboratory scientists are able to consider which variants are shared in the family, and who is affected, to arrive at this much smaller pool of variants of interest.

Next, information about the patient’s phenotype – their clinical presentation – is used to determine which of these could plausibly be the cause of the condition…

Once those variants that are not thought to be linked to the clinical presentation (lime green) have been discarded, there will be an even smaller pool of variants to look at in more detail.

As these remaining variants are in genes or areas of the genome that are all considered to be closely related to the clinical presentation, they require expert analysis…

The final stage of variant filtering, therefore, is not part of the bioinformatics pipeline but is done by clinical scientists, who have the expertise to assess the likelihood that each of the remaining variants is responsible for the condition in question. This often involves collaboration between clinical scientists and other experts from different disciplines as part of a multidisciplinary team (MDT), as well as the use of numerous tools and databases.

At the end of the process, the aim is for clinical scientists to be able to determine which variant, or variants, are responsible for the disease in question.

Whatever their findings, the clinical scientist will produce a report so that this can be fed back to the patient and family. Increasingly, a diagnosis can inform treatment and management decisions.

Ordering a test: What you need to know

Phenotype
  • It’s vital that you provide as much detailed information about the phenotype as possible – even if you’re not sure whether it’s relevant. Even a small omission of relevant information could alter the filtering process and mean the variant of interest is filtered out during one of the filtering stages outlined above. In addition, the handful of variants left for the clinical scientists to analyse at the end of the machine filtering process may be interpreted differently in light of the phenotypic information provided, so it’s vital that it is detailed and precise. If in doubt, include it and don’t disregard it. 
  • If your patient’s data is to be analysed using virtual gene panels, as the clinician who has examined the patient you should select the correct panels upfront. It is often difficult to adjust the analysis later down the line; getting the analysis set up correctly at the start maximises the chance of a diagnosis. If you aren’t sure which gene panels you need, it may help to contact your laboratory or specialist colleagues in clinical genetics for advice.

How is the information used?

Information about a patient’s phenotype is the basis for deciding which genes and areas of the genome to look at following sequencing. Phenotype information can feed into the analysis in three different ways:

  1. Bioinformaticians can run an automated comparison between the patient’s condition (expressed in standard terms such as those used in the Human Phenotype Ontology, or HPO), and the genomic variants found in that patient.
  2. The genes most likely to be relevant to the patient can be identified up front by the patient’s clinician, and used to direct which variants will be prioritised during variant filtering. For efficiency and in an effort to standardise this process, this is often done using virtual gene panels – lists of genes which are known to cause specific conditions such as cardiomyopathy or dystonia. The NHS records its gene panel content in PanelApp.
  3. During the final stage of analysis outlined above, clinical scientists will compare what is known about a gene or variant found in the test with the information they have available about the patient’s phenotype.
Parents and family
  • If it’s possible to obtain DNA samples from, and health details about, both biological parents, it can be useful in some cases. This is especially true when testing children or young people for complex developmental disorders. For adults with later onset conditions that are often passed down in families, it is better to test only your patient to start with (as long as they have the familial condition).
  • If a child and both parents are included in the test, the analysis will focus first on variants that the child doesn’t share with either parent (de novo variants). These are a common cause of genomic disorders, and having the family information makes the analysis process much more robust and straightforward.
  • There are risks of comparing sequence data from different family members. For example, if the child and the mother are thought to share the same condition, only variants the child has inherited from the mother will be considered. If in fact the child has a different condition from the mother, you are likely to miss the diagnosis with this analysis.
  • Genomic conditions can affect different members of a family in different ways. Some disease-causing variants can be passed down in families, but for reasons we don’t fully understand, not all genetically-affected individuals develop symptoms. If you are unsure about the way a disease is affecting a family, it is likely to be best to start by testing just one affected individual. If you need advice, you can contact your local laboratory or experts, for example, in clinical genetics.
  • In some cases, and for a number of reasons, access to both parents and their health information isn’t possible. In this instance, there will be more variants for the clinical scientists to analyse at the end of the interpretation process, which will normally take longer. In some cases, this may mean that the variant of interest cannot be found because there are simply too many to analyse. A strong phenotype is really helpful in this case!
Interpreting results

As explained above, the final stage of variant filtering is always led by clinical scientists, in consultation with clinical colleagues as part of the MDT. As a referring clinician you won’t be directly responsible for variant interpretation, but there are a number of very important considerations that you should bear in mind:

  • The diagnostic report relates the findings of the test back to the individual patient or family, whom the scientists will not have seen. You may be asked by the clinical scientist working on the results to contribute information or an opinion, either within an MDT meeting or remotely. For example: Is it possible that the mother could be affected with a milder form of the patient’s condition? Is there any family history suggesting an X-linked inheritance pattern? Does the patient have microcephaly or seizures? Your swift contribution is invaluable.
  • A high level of confidence is needed when it comes to interpreting genomic variants – i.e. deciding whether a variant or variants identified through genomic sequencing are the cause of disease. This needs to be based on appropriate evidence, under expert oversight, as Dr Richard Scott and Dr Angela George explain in our videos. The NHS has processes in place to ensure interpretation is as accurate as possible, drawing on scientific and clinical expertise both nationally and internationally.
  • Results may sometimes be issued initially prior to a formal report being provided, for example during MDT discussions, or to help with very urgent clinical situations. In general, a formal report should be available before a result is discussed with patients or used for clinical care.
Genomic reports

Once the filtering and interpretation processes are complete, a genomic report will be issued by the clinical scientist:

  • If a pathogenic or likely pathogenic variant has been identified as the cause of the patient’s condition, this result can be used for clinical care; to direct patient treatment, offer testing to other affected or unaffected family members, and/or to inform reproductive choices.
  • If a variant of unknown significance has been identified, the result should NOT be used for clinical care. However, there may be further information you could help with, such as obtaining samples from wider family members, which could increase confidence in the variant being relevant or irrelevant for the family. This will be written in the report.
  • If any of the information in the report is incorrect, for example if the phenotype information is incomplete, or the disease status (affected/unaffected) of family members is incorrect, contact the laboratory to let them know as soon as possible as this may have affected the interpretation and reporting of the result.

Summary of key points to consider

Requesting a test: FAQs

1. How do I select patients for genomic testing?

The National Genomic Test Directory for Rare Disease specifies all the available tests and which patients are likely to benefit from a test. In general, the directory focuses on the patient’s context for all tests: If a disorder is likely to have a single genetic cause, and the patient needs a molecular diagnosis to direct their clinical care (or that of their family), then in most cases they will be eligible for a genomic test.

2. What are HPO terms and how do I provide them?

HPO is the Human Phenotype Ontology. It is a standardised way of – a language for, if you like – capturing phenotype information such that it can be used in an automated analysis. There are a number of websites where you can search for the relevant terms to describe your patient’s phenotype, including the one linked above and this one.

Providing HPO terms for unaffected relatives may also be helpful. If the relative’s disease status is listed as unaffected then the analysis will not be limited to variants that relative carries. However, if they have a mild phenotype it may be useful to use this information in the interpretation process.

3. Why do I need to select gene panels for genomic tests?

For some bioinformatics pipelines, particularly where a patient is being tested without their biological parents, it is important to set the appropriate analysis target up front, to reach a manageable number of variants for the final manual review. This is the responsibility of the requesting clinician. NHS panels can be found here, using PanelApp.

If a patient has a complex phenotype, they may need several panels applied to ensure all relevant genes are examined. For example, for a child with developmental delay, regression, microcephaly and seizures it might be appropriate to specify the panels ‘intellectual disability’, ‘genetic epilepsy syndromes’, ‘inborn errors of metabolism’ and ‘mitochondrial disorders’. In general it is more difficult and less efficient to expand the target for analysis after the test has taken place, so providing this information at the pre-test stage is optimal.

4. Should I always try to provide samples and data from other family members?

No. For a child or young person with a complex developmental disorder, providing samples and data from both biological parents increases the chance of making a diagnosis and improves the efficiency of the analysis. For adults with later onset conditions that are often passed down in families, it is better to test only your patient to start with (as long as they have the familial condition).

Providing a broader family history may also be helpful, for example if there is a history suggestive of an X-linked condition.

Additional family samples may be needed after the initial result is available, to check whether a variant of unknown significance is tracking together with the disorder in the family.

5. What is penetrance and how do I know which setting to choose?

Some disorders have ‘complete penetrance’ – that is, a person who carries the disease-causing genomic variant will always develop the condition (sometimes only at a later age: age-dependent penetrance). Other disorders show ‘incomplete penetrance’. This means that some people with the disease-causing variant may never develop symptoms of the condition.

Some analysis pipelines can be set to assume that penetrance is complete or incomplete for a specific analysis. Unless it is very clear that penetrance is highly likely to be complete, it may be best to assume incomplete penetrance, as this covers both possibilities.

6. What do I need to tell my patient about genomic testing?

We have developed a range of resources designed to support health professionals with offering genomic testing to patients.

You may also like to check out our video series ‘Let’s talk about… genomic testing‘.

7. What samples are needed for genomic testing?

Most genomic tests require a blood sample. The sample type is often specified on the test request form. For some specific test types a different sample may be needed. If you need more information, please check the National Genomic Test Directory, or ask the laboratory.

8. Where can I find more resources for facilitating genomic testing?

Check out our web page about the National Genomic Medicine Service, where you’ll find lots of helpful information, online courses and competency frameworks.

9. How do I find my local laboratory?

A full list of the Genomic Laboratory Hubs can be found here.

Many of the hubs also have their own websites where you can find more information, including contact details and how to order a test:

Why might we not have an answer?

Unfortunately, genomic testing may not always provide an answer to explain the clinical situation. Why?

‘Big data’ challenges

Our understanding of genomics is only as good as our data pool.

One of the early filtering stages shown in our image carousel involves throwing out all the common variants identified by sequencing. But, as we obtain more data, from a larger proportion of the human population, we will learn more about human variation. This will change our interpretation of genomic variants over time. For example, a variant that currently looks very rare, which is therefore a candidate for diagnosis, might turn out to be relatively common in a part of the world where there is currently little sequence data, and therefore in future will be removed during filtering. 

Limitations in knowledge

There’s lots we still don’t know…

In addition to encompassing the full range of human diversity, more data will also enable us to understand more about both the protein-coding and non-protein coding regions of the genome, which will greatly advance our understanding of genomics. For example, if the variant we are looking to find for a particular patient – our pink ball – is in a non-coding region of the genome it will be disregarded by current testing pipelines, because we are currently limited in our understanding of these kinds of variants. In future, when we understand more, genomic sequencing may include such variants – and an answer may be found where it may sadly not be today. Read more about the answers that may lie in non-coding DNA on the Sanger website.

Note: Occasionally, new knowledge may show that a variant we previously thought was likely pathogenic or pathogenic is actually a benign variant which isn’t causing disease. Because of the high evidence threshold for using a variant in clinical care, this happens very rarely. If it does happen, the laboratory will aim to issue new corrected reports and return these to the relevant clinicians. Specialist colleagues – for example in clinical genetics – may need to support the process of communicating this to patients.

Technological limitations

Genomic sequencing doesn’t currently look at all the changes in the genome.

Some potentially important variants in the genome may never make it into our pool in the first place. The balls in our initial pit are those that are currently confidently detectable using standard sequencing technologies, but each type of genomic test will have technical limitations which mean that some genomic information may not be captured, and therefore biologically relevant changes may be missed. As techniques improve, so will the ability to capture these types of variants.

As an example, triplet repeat, or STR variants, which cause some neurological conditions, are not detectable on exome or panel sequencing tests; though many are now detectable by whole genome sequencing. Conversely, medium-sized sections of missing or extra DNA (deletions and insertions of around 50 to 2000 DNA base pairs, or letters) may be detected on panel or exome tests, but less effectively by whole genome sequencing. The genomic report may include specific information about what variant types have been included in the analysis.

There are also some complex genomic situations which may be harder to detect. For example if a patient is mosaic for a genomic disorder (only some of their body’s cells are affected with the disorder), this may be missed by standard tests.

There is no identifiable genetic diagnosis

The condition might not be genetic – or it may be too complex for our current understanding. 

Some families may have a condition that seems to run in the family but doesn’t quite follow the ‘rules’ we would expect for a condition caused by a single gene variant. These conditions are often caused by polygenic or complex inheritance – in other words there are a large number of variants each contributing a small risk of the disease, rather than a single high-impact variant. At present, we are unable to offer clinically relevant genomic testing for these conditions in many cases. If the condition is thought to be caused by a single gene and a genomic test is arranged, the result will be negative as there is no single variant causing the condition.

In some cases, a condition may have a completely non-genetic cause. It is sometimes hard to tell the difference between genetic and non-genetic conditions as patients may have very similar phenotypes. If genomic testing is performed for a patient with a non-genetic cause, the result will be negative.

Individual circumstances

It may not be possible to make a confident diagnosis.

There are some situations where interpreting genomic tests may be difficult, for example in a young baby who may not yet have developed recognisable features of a condition, or in a non-specific presentation such as advanced renal failure. In these circumstances, the clinical presentation may develop over time, in the original patient or in their family members, and in this instance it can be helpful to revisit the diagnostic question in future.

Towards genomic equity

Hear from bioinformatician Nana Mensah about the current limitations of our data – and why it’s so important to address it.

Read more

Further learning

This site uses cookies. By continuing to browse the site, you are agreeing to our use of cookies.

OKLearn more