Showing posts with label Google. Show all posts
Showing posts with label Google. Show all posts

Tuesday 19 September 2017

Identifying your sub-type of Autism

Today’s post is very much a work in progress, so do not expect all the answers.
It has occurred to me and also some readers of this blog that you could produce a diagnostic decision tree that would narrow down each person’s sub-type of autism. This really does lend itself to a relatively simple computer model, meaning you could have a simple on-line diagnostic program/app. You do wonder why a tiny part of the hundreds of millions of dollars spent on autism research is not allocated in this direction.

A blank screen awaiting more case studies

A company called Verily, formerly Google Life Sciences, would be an appropriate partner for such a project.  Google is currently backing the approach that genetic testing will reveal all about autism, which looks unlikely.
The decision tree computing part of such a project could be done in an afternoon, what would like more time is creating the logical diagnostic steps.
One clever part, where Verily could help, would be the use of software to “read” the research and look for associations; this is after all what I do with the help of Google Scholar and Ctrl F. In this way you could quickly “read” many tens of thousands of research papers, books and case histories, if there were any.
Software can also be used to “read” all the personal data of any volunteer participants, such as genetic tests, MRIs, family history and all lab tests. As our reader Tyler pointed out, just relying on a pair of eyes to read the MRI means tiny variances go unnoticed and perhaps they do mean something.
You cannot automate everything, but technology certainly can help.
Even a primitive decision tree can help, because it pushes you to think logically and stops you wasting time potentially applying therapies that might help some biological dysfunctions, but clearly not yours.

Parent led initiatives have been successful in the past and have resulted in rare diseases or syndromes becoming treatable using so-called orphan drugs. Treating autism is a massive task and parent led initiatives are what exist today - they have not succeeded in making any impact on mainstream medicine. To make an impact in autism you need big money and big companies.

Start with severity
Ideally everyone would have autism diagnosed in early childhood.  In the case of mild autism this usually does not happen.  Ideally all children would be assessed using the same evaluation test, like CARS, but this also does not happen; most people have no evaluation test, just a subjective observational diagnosis.
So you would need unambiguous terminology to define how the person is affected by “autism”.  This would be how verbal they are; level of cognitive dysfunction and how “autistic” they are behaviorally.
Then you can add things like epilepsy, self-injury and adaptive behavior (toileting, feeding etc). 

Family History
While genetic testing is seen by many as the ultimate diagnostic test, in many cases family history tells you a great deal more. Similar family histories will very likely end up with similar sub-types of autism. This would include medical history of relatives, but also educational/job attainment. It would include health before and during pregnancy, type of delivery and health status at birth.
Genetic testing of your DNA can never be the holy grail of autism diagnosis; what matters is gene expression, when and where, and this is something much more complex.
Family history tells you what has happened in the past and most things happen for a reason.
Family history is by its nature very personal, but it is one of the richest areas to identify how people might be usefully grouped together. This may appear highly politically incorrect, but that does not stop it being useful. 

Physical Features
Until very recently even single gene types of disorder were actually diagnosed by their hallmark physical features. These are often facial, or on hands or feet, but can be anywhere.
One of the most useful physical markers is very simple, you just need to know at birth and for the next couple of years was there a tendency towards being big and/or muscular or small and more floppy?  The stronger the tendency to either extreme, the more important is the observation.

A recent study has even used 3D modeling to identify autism by analyzing people's faces. This apparently works particularly well in females with autism, who apparently tend to have more male features. I have no doubt Googlers could improve this.

Hypermasculinised facial morphology in boys and girls with Autism Spectrum Disorder and its association with symptomatology

Other medical conditions in the child and also in the parents, and in particular in the mother during pregnancy, can help assign people to subgroups. 
In mothers thyroid disorders and diabetes during pregnancy look particularly relevant.
In the child, epilepsy, allergy and GI issues are important. There are many types of GI issue, some of which overlap with allergy and some do not.
Then you have sleep disorders, eating disorders, sensory disorders etc.  Some are only an issue when very young and then fade away, some may remain. 

Biological Markers
There are hundreds of possibly relevant biomarkers and many more that are likely totally irrelevant.  The most reliable tests will use samples taken from spinal fluid, which in effect is part of the central nervous system, and so tell you what is going on inside the blood brain barrier. Blood tests can be useful but often do not tell you what is going on inside the brain. 

The most immediately relevant biomarkers relate to treatable in-born errors of metabolism, each one of these syndromes may indeed be rare, but there are many of them and you would think it worth ruling them all out. 
The risk here is getting lost in hundreds of tests, rather like with genetic testing.
There will be some useful markers like for oxidative stress, inflammatory biomarkers, mitochondrial function etc. 

It does make sense for everyone with autism to have an MRI, including those with Asperger’s. People who break their arm get an X-ray by default, is an MRI for autism too much to expect?
The brain may appear entirely normal, but there is a significant chance of identifying a relevant variation, albeit small.
There are also some specific tests that can be carried by Functional Magnetic Resonance Imaging (fMRI) at the same time as the basic MRI.  

Many people with autism, but without epilepsy, do have an abnormal EEG, with so-called epileptiform activity.  Medicine does not have a consensus opinion regarding what to do, but some neurologists, like Dr Chez, believe it should be treated and can reduce the severity of autism.  Detailed knowledge of such epileptiform activity might be useful when defining sub-groups of autism.

Genetic Testing
The tests usually offered are microarray, whole exome and whole genome sequencing.  These tests may reveal something useful or may just reveal a list of variances that do not seem to have any relation to autism.

If funds are unlimited, then whole genome sequencing of the child and both parents is the best choice.  Then you need someone very methodical to review and interpret the results.

What matters is gene expression locally and even whole genome sequencing does not tell you this, it is however a part of the picture.

Nature of Onset
People tend to consider autism as either early onset or regressive. Here it is important to extract the group who were born entirely neurotypical, met all their milestones and then regressed.
Most people’s autism started long before birth but it manifests itself differently over time. This is like the progression of a disease. Indeed one therapeutic idea is that by very early pharmacological intervention you can affect this progression and improve the final outcome.
The point here is that you can have early onset autism that appears to “get worse”, this is not regressive autism.
People with regressive autism need to split into those who profoundly regressed and those who were always different and then became more so.
For people with mild autism the time/nature of onset is likely much less of an issue. 

Variability of Symptoms
Some people with autism exhibit the same severity of symptoms every day, but many do not.  In people with highly variable autism it is very useful to understand what makes it worse.  Many lay people refer to autism as being the result of the brain being "wired-up" differently, as if it was static condition. In those with highly variable autism, the condition is clearly not fixed. Something is making autism worse, you just have to find out what it is. It could very likely be allergy, but could potentially be many things even microorganisms affecting gene expression.

The Goal - Why would you want to collect all this personal data?
As I have noted in this blog, there clearly are groups of people who respond to the same therapies. It is remarkable and it is not a matter of chance.  I do not just mean one therapy, like NAC or bumetanide, but a whole string of them, even the ones that may appear odd to some readers.
If you could predict who fits into which subgroup of autism based on answering a long list of questions on an app on your smartphone and sending some data to Google, widespread treatment of autism would become a reality.
You do need a lot of data and most of it does not yet exist.
I know a lot about one subtype of autism and there is a substantial overlap with the son of one regular commenter. Bumetanide, Potassium, NAC, Atorvastatin, Verapamil, Potassium Bromide etc all appear to have the same (beneficial) effect.
Once you have more data, you look what all these children have in common and then you can try and predict which other people will have a similar drug response. 
You just need a lot of well documented case studies and a lot of experimentation with possible therapies. In other words not just Peter’s Polypill, Tyler’s Polypill, Alli’s Polypill or Dr Kelley’s mito-cocktail,  but very much more.
There does seem to be an effective therapy for regressive autism caused by mitochondrial disease, which is likely to be one of the larger subgroups. Given all the attention given to PANDAS/PANS, I cannot understand why the mitochondrial sub-group has not been similarly “mainstreamed”. The therapy is Dr Kelley’s (from Johns Hopkins) plus add-ons.  One potential add-on being calcium folinate (Leucoverin) to reduce peroxynitrite (ONOO) from nitrosative stress, but there are undoubtedly other add-ons waiting to be discovered, or perhaps just documented.


Trying to match effective drugs to a long list of markers and other criteria, does have something in common with tracking individuals web browsing to generate personalized relevant advertising. So I do think that the likes of Google/Verily would do a much better job than medical researchers alone. They would of course automate the process, which is not what many doctors are going to like. 

If you happen to know Larry Page or Sergey Brin from Google, you can suggest it. Brin's former wife incidentally is the founder of 23andMe. I am sure both companies employ plenty of people with Asperger's and so likely have some kids with autism. Brin apparently is interested to find a Parkinson's cure, he has a mutation in the gene LRRK2; since this gene is also associated with Crohn's disease, he might want to fund that too. Why does Brin have a LRRK2 mutation? You only need a glance at his family history.