An Easy Kludge to Examine SNPs

It turns out that there is a large effort to make available all sorts of information about genomics, especially in the medical field, for free including multiple tutorials and results of government financed studies. 
One of the problems for the clinician is that there is too much out there and there are at least three different sets of words for everything studied because for a long time the geneticists, clinicians, and gene researchers didn't talk to one another.  For example, rs1800044, a SNP variant found in a patient with Tourette Syndrome is also known as 657G/T or Arg219Leu and never shall the twain meet, at least in the early research papers.  (Here is a paper that does mention alternative names.)
So if you look up the research sequence number you might not find an important paper under that name and most likely won't know the other nom de genes.
But clinicians don't need to know all the technical aspects of the genome, they need to know how to use the information and what information is both pertinent and valuable for patient care.
Unfortunately most of this information is still hidden in the bowels of research labs and obscure journals.  I suspect that in five years there will be a whole industry whose sole purpose to to sell you software that helps find this information and that the drug companies will be very eager to give you the information that helps sell their products.   The recent brouhaha involving 23andMe and the FDA will most likely be resolved in favor of the money meaning that eventually everyone will have access to the information without oversight of the FDA.
The fact is that anyone with $99 can get this information right now, at least that information found in the exome, that part of the gene set that makes the majority of proteins including enzymes and receptors.  A vast number of SNPs correlated with disease states and medication response are included in these surveys so it is just a matter of getting the information and the individual exomes together is a way that helps the clinician and the patient.
There is a kludgey solution that involves a curious clinician, a computer, Firefox plus an add-on called SNPTips, and a website named SNPedia. 
The solution is a kludge because it involves a Rube Goldberg connecting of the various parts and because it is incomplete.  Not all the information known is on SNPedia - the research is coming fast and furious and is hard to keep up with and this is a wiki effort - and they only have slightly over 50,000 SNPs cataloged at present.  Nonetheless SNPedia is a terrific source for information.
The other reason that the solution is a kludge is that there is no good way to vet a lot of this information.  For the most part the studies are small and focused narrowly on one geographic group.  Hence what is true for a Han population may not be true for a European population in any given instance due to actual differences or the weak power of the studies.  On the other hand, there are some very large studies that are helpful and intriguing results are coming out all the time.
As clinicians we are primarily interested in helping our patients get better.  In psychiatry this usually means making them feel better without causing a lot of other problems such as tremendous weight gain, debilitating side effects or transitioning them into another phase of their illness. Sometimes we have to struggle with the diagnosis or rethink our strategies with patients who fail in treatment.
Genes account for some percentage of the illness we see in individuals.  The days of nature vs. nurture are pretty much over as we have come to the realization that both are involved in the disease process.  The recent upsurge in genomics research that followed advances in gene assays has resulted in a much more sophisticated understanding of how genes work.  For most of us older physicians the language and information is overwhelming but with a little study (and a tolerance for boredom) you can  glean enough information to help use what is available.
I won't go into how to study up on genomics in this column, but I will tell you how you can adapt your existing computer to help you learn more about your patients.
I am an Apple computer guy.  That's because I move my lips when I compute and have found the more expensive alternative easier to use.  But what I am going to say is equally applicable to Microsoft or Linux operating systems.
The secret is to get Firefox from Mozilla and put in an add-on called SNPTips.  SNPTips reads any research sequence number (the "rs" number) and refers it to SNPedia.  The raw data from 23andMe is in text form and easily read.  (I am sure that there will be others offering sequencing with the same data soon.)  When you click on the highlight generated by SNPTips, you are shuttled over to SNPedia which in turn will refer you to the relevant research papers. 
The opposite can also occur.  SNPedia allows you to look up various topics such as medications or diseases and lists the SNPs that are involved with those subjects.  SNPedia is not complete, it is a wiki based entity that relies on the kindness of others, but it is very assiduous about the completeness of the information that it has so it is a good resource.
Here are the steps required to do this:

1) Go to Firefox and download the web browser.
2) After you have started Firefox, go to SNPTips and install the add-on.
3) Have the raw data from 23andMe available on the computer somewhere.  I use a thumb drive in a USB port with an anonymous file name.  Go to preferences in the SNPTips add-on and browse for the thumb drive. 
4) Look up the SNPs you are interested in anywhere on the web and the SNPs will highlight in green if they are on the raw data. By placing the pointer on the highlighted area will tell you which allele is present.  rs6280.  I just put that in there to help you see if you have done everything correctly.
5)  If you click on the highlighted area it will take you to SNPedia for further information. 

What I would like to do with this blog is develop (with the help of others) small disease or medication related pages that can be used as a tool by clinicians.  There are a lot of ethical and medical issues that have to be addressed and resolved but I think that the system I am outlining can be kept private and off of the internet by having the information anonymous and local and always in the possession of the patient until there are laws and software protecting access to this information.
For now it is in a proof of concept stage.  I will provide a few anonymous raw data sets that have been modified from the originals for practice and make them available to others to use as part of this project.
At some point a more sophisticated solution will become available, I am sure.  It will require a combination of clinicians learning more genomics, patient cooperation (23andMe is cheap but it is still $99) , and a thorough evaluation of the value of such information.  We are on the cusp of this technology breaking out and it would be very helpful to be able to use all of the free information that has been made available.

Michael Keyes MD
Fond du Lac, WI

The Olanzapine Example

Suppose you have a device that will not only tell you the diagnosis but also tells you the best treatment for a patient you have never seen before.   I suspect that it would be very handy, especially in a busy practice.
Now suppose that not only does this device tell you about the medication needed but it also helps determine if the patient will suffer from a significant side effect that may permanently cause physical damage. 
Lastly, this device does not cause pain nor does it require blood, just saliva.  Imagine how useful that would be.
In the real world, such a device would take years to develop and millions of dollars to become a reality.  It would also require significant time and money to pass the FDA barriers in place to protect society.
On the other hand, if you could elicit information from the patient that did the same thing via your interview and observation, the cost would be minimal.
This is the dilemma of genomics in medicine.  The information this there, it is 98%-99% accurate, and it is cheap.  The research for a lot of the uses of this information is still in its infancy but there are some papers that are well researched and practical that are not only intriguing but probably very useful if we could find a way to utilize them.
One example are the genomic studies that involve the anti-psychotic olanzapine.
Olanzapine was first introduced as Zyprexa in the United States as an atypical anti-psychotic that also has mood regulating capabilities.  It went generic in 2011 and is a standard drug for psychotic illness.  The drug has numerous side effects including massive weight gain and a proclivity towards hyperglycemia.  It has proven to be a useful drug but the weight gain issue often requires that the patient switch to another drug before permanent problems occur.
There have been a number of studies on the relationship of certain SNPs to olanzapine in regard to   weight gain and effectiveness.  A study of 88 subjects showed that those with the ancestral variant (C;C) in the rs6280 SNP were more likely to respond to olanzapine that those with the T;T variant.
A study of 67 subjects with the ancestral variant of rs7412 (C;C) who  took olanzapine showed significant weight gain as opposed to the T;T variant.  Two other SNPs, rs5092 and rs4765623 were also implicated in weight gain with olanzapine.
This sounds great, you might think, but there are significant limitations to these studies that need to be clarified.  First, the numbers in these studies are very small.  This could easily been an outlier and a study with thousands of patients might come up with a totally different result.  Second, this is a study of a single group, Caucasians in this case, and we know that various geographic groups have different ratios of responses to a variety of medications.
Third, genes are not disease.  There was a significant tendency for those patients with the variants mentioned to gain weight so there is a risk factor that has to be taken into account but it is not 100% predictive.
Nonetheless, if you have a number of medications you can use that appear to be of equal value, it probably would not hurt if you looked at these SNPs to help you decide whether or not to use olanzapine.
You can bet that if we had the technology in 2007 (the date the paper was published with government money) that Eli Lilly would have had a med rep on your doorstep touting Zyprexa and offering free 23andMe assays.  I understand that it is becoming a standard part of the testing for some researchers to correlate SNPs with their subject drug.  Gene-Wide Association Studies are becoming more available both for drugs and diseases these days.
So if you have a patient with the 23andMe assay, a computer with access to SNPedia, Firefox and an add-on called SNPTips, you could look this up very easily and be able to use the information to make a decision about the use of Olanzapine.
For instance, my rs6280 is T;T, the variant that is less responsive and my rs7412 is C;C correlated with weight gain.  This does not mean that I will have a poor response to the medicine and grow immense in the process, but the likelihood of both is much higher than if I was on another anti-psychotic.
All of the SNPs I mentioned happen to be in the exome analysis from 23andMe (and other vendors, they all use the same lab/chip) which makes the studies available helpful allowing for the warnings about power and scope mentioned above.   Olanzapine/SNP correlation is only one example of what is available so long as you know to  take it with a grain of salt.   Because this type of research is getting cheaper and easier ( GWAS  is often done with available data bases, no humans were harmed in these studies) we will see more and more possibilities like this to help our patients.
Mayo Clinic, Vanderbilt University, and a host of other well known medical centers are already employing these tools in order to individualize medicine.  Many of these centers are still using expensive specific studies but it appears that there are cheaper alternatives that are as accurate, at least as far as the SNPs are concerned.   The trick is to get the analysis correct and that will take some time.  I suspect that that time is short, however and learning to find your way around the confusing genomic landscape is not all that difficult if you work at it a little.

Here is a little cheat sheet for olanzapine.

The Olanzapine Cheat Sheet

Here is a quick précis of the SNP/olazapine relationships.

rs6280 (the “rs” stands for research sequence.)

The ancestral allele is C;C (cytosine.) If a patient has C;C then they are more likely to have greater remission of positive symptoms of schizophrenia.

If the allele is C;T or T;T (thymine) then the response is poorer. This is because there is a substitution of another amino acid on the D3DRD3 receptor.

Rs7412

The ancestral allele is is C;C. Those with the ancestral allele are more likely to gain weight on olanzapine than those with the T;T variant.

So, if you have a patient that has the C;C on the rs6280 he or she is a good candidate for olanzapine (with the usual warnings about the power of the study) but if he or she is C;C on rs7412 then they might gain a lot of weight.

Ideally the patient is a great candidate for olanzapine with rs6280 (C;C) and rs7412 (T;T)

Michael J Keyes MD
Fond du Lac, WI

Proof of Concept

Psychogenomics is a blog that is designed to help with a proof of concept project based in the Behavioral Health Section of St. Agnes Hospital, Fond du Lac, WI.
The idea is to see if readily available genomic information from 23andMe or similar companies can be transformed into information that is useful for both the patient and their provider in regard to treatment and diagnostic decisions.
In the past ten years or so there has been an exponentially growing body of knowledge regarding genes and disease.  The field of psychiatry has been part of this epiphany especially in regard to how medications affect both the disease processes seen in patients and how patients respond to medication.  In addition there has been work linking these genes, especially the Single Nucleotide Polymorphisms (SNPs) that act as markers, to diagnosis.  The latter research may have global influence on how we view diagnosis in the near future.
Mental health providers rely on the Diagnostic and Statistical Manual 5 (DSM-5) as the common nomenclature of psychiatric disease.  This work is not based on strong scientific bases, rather it is reliant on describing and categorizing phenomena according to both historical and clinical observations.  The book takes clusters of symptoms and signs and assigns diagnoses based on these often overlapping conglomerations and then assigns a name.  In turn theses diagnoses are classified into larger units based on mood, thought processes, cognitive abilities and a host of smaller categories.
The DSM-5 is not close to perfect, but it is all we have and it serves as a rough guideline for treatment.  Ironically, many of the treatments are capable of crossing the diagnostic lines and help patients with varying diagnoses.  Thus a patient with schizophrenia may end up taking the same medication that a patient with a depression does and both patients improve.
I say improve because, for the most part, psychiatric illnesses are chronic and incurable.  The major psychiatric illnesses - Major Depression, Mood Disorders such as Manic Depressive Disease, Schizophrenia, Schizo-affective Disorder, Anxiety Disorders, etc. - all seem to be inherited although there is hardly a one to one correlation to family history and the presence of illness.  Patients who have demonstrated severe illness probably have a genetic component that will not go away with the treatments we have now.
These days most providers realize this and have a host of tools (medication, therapies, somatic treatments) that will help but because of the fuzzy nature of diagnosis, the vast array of medications and treatments, and the significant side effects that can occur, especially with medication, it is often a crap shoot as to what will help and what will not.
There have been a number of large open studies such as STAR*D for depression that looked at tens of thousands of patients and developed algorithms for treatment.  Instead of just tossing one of sixty or so antidepressants at a patient to see it it works, the STAR*D study came up with a flow chart that allowed you to treat in a "logical" manner according to the statistical analyses of the study.
While these algorithms are very helpful (they exist for mood disorder and psychoses too) they are still not as precise as they could be.  Some of the reasons for this imprecision lies in the fact that individuals have specific responses to medications based on their enzyme efficiency (i.e. how they metabolize the various medications) and their susceptibility to illness.  There are other factors, of course, but these days we have the technology to determine what enzyme variants a patient might have and, if the studies are correct, even be able to tell if a specific medication will be more appropriate for a patient.  In addition there is some proof that we can tell what the disease processes is (remember, a lot of them look alike) through a study of the patient's genetic structure.
Thus this project. 
23andMe makes available to consumers a snapshot of their genes from a part called the exome.  The exome is responsible for the production of  a significant number of enzymes and the readout from 23andMe contains about 1,000,000 of the SNPs mentioned above.  (For contrast there are probably over a billion SNPs in the body.)  Of these million SNPs there are a finite number that are involved in the production of important enzymes and medical science has identified a large number of them.  SNPs are segments of the genes involved and they can have variation in the amino acids that make them up.  When these SNPs change - and the changes are limited to only four amino acids - they effect the ability of the enzymes to work or have some other not so obvious change in the whole system.  Each of us carries a pair of these alleles so it is possible to have at least four variants and as many as seventeen in one case depending on whether or not the SNPs are duplicated, repeated, or even there (there are some people who have certain SNPs missing.)   It can be very confusing especially if SNPs influence the effects of other SNPs but on a practical basis most of the science is moot as we tend to look at the results and not the biochemistry.  (Just like the rest of medicine, it is an applied science, after all.)
So if this information is available, and chances are that many of the million SNPs are available for the $99 that 23andMe charges, we should find a way to utilize it.  It appears that we are able to access these SNPs and the way is rather inexpensive because a provider can use open sources while retaining the confidentiality demanded by ethical practices and the law.
Thus I am setting up this blog.
It will serve as a test bed for ideas and can also be used to generate ideas through crowd sourcing.  Several friends of mine have anonymously donated their  23andMe raw data for this project and St. Agnes has generously donated money and expertise for the computer part. My main source of inspiration is the book "Psychiatric Pharmacogenomics" by the late David Mrazek, MD who, until his untimely death, was chairman of the behavioral health section of the Mayo Clinic.
I guess we will see.

Disclaimers:
1) This blog is not intended to diagnose or treat any medical illness.  It is merely a meeting place for ideas that will help the professionals become more precise and accurate.  (There is a difference, look it up.)

2) I am not endorsing any product mentioned here and St. Agnes Hospital certainly isn't endorsing any product.  They are just providing me with some  support because they think I may have a good idea.  Vendors such as 23andMe will be coming out of the woodwork once the FDA business is over because they provide an accurate service that defines SNPS.  The books mentioned are considered to be important in the field but are rife with controversy according to some.  If you have a better source, please let me know.

Michael J Keyes, MD
Fond du Lac, WI