Participate in a Pilot Study of Genetic Patterns in PANS/PANDAS

[bobh]

You are invited to participate in a research study titled “Genetic Patterns Among PANS/PANDAS patients”.

This study is being conducted by Bob Horvath, Mark Moeglein, Michaela Holden and Sam Keating. We are "citizen scientists” with some qualifications in statistics and data manipulation, and direct experience (ourselves or family members) with various autoimmune or immunological conditions and/or autism, all of which we suspect are related.

The purpose of this study is to find out whether certain variations in DNA occur more commonly among PANS patients than the general population. If we find some, we will let you and many others see the (aggregate) results, and hope that a finding will spur further research into the biological processes, and ultimately, in the long term, possibilities for remedies for the disorder.

Participation in this is entirely voluntary. You can choose not to participate at all, or to participate anonymously, or with your name attached to the data. Either way, there will be no effect on your relationship with the researchers, or any other negative consequences with not participating.

You are being asked to take part in this study because you are a parent of a PANS/PANDAS child, or you are an adult that has suffered with PANS/PANDAS symptoms as a child, or as an adult.

If you agree to participate, you will be asked to click on a link below and upload 23andMe or similar genetic data for one person (at this time, please don't upload multiple family members, just the one).  You may also email your data (see below).

The data will be collected regularly from the upload site, with a January 19th, 9pm EST cutoff time for data used in the study. Data uploaded after the cutoff time will be safely stored (with no direct identification of participants) on two computers only, for possible later confirmation of any result with the initial data. Even if you give your name, the data will not contain that name, but be coded before being processed. After the initial upload and the coding step, no other person, website or online service will have access to your data with your identification attached to it. Your data will be uploaded to GEDmatch (without your identification) in order to obtain ancestry, and to confirm no close relationships (>3%) to other participants. No genetic disease information will be extracted from participant data.

For those that contribute anonymously, the only link between us and you will be a fake name and email address that you give at the upload site. You are free to withdraw from this study at any time. However, once you submit your data, the only way to withdraw your anonymous data is if you contact myself (Bob Horvath) and reveal your fake name, so that I know which data to remove. This step could reveal your identity, but your data will be removed from the study.

Study data will be stored without names in digital format. One copy will be on a computer in Ontario Canada, and another will be kept on a computer in Oregon, U.S.A. Only Bob Horvath and Mark Moeglein will have access to this full data. Aggregate results for the analysis of the data will later be made known publicly.

There are no known risks associated with this study, beyond any risk there may be associated with the data existing (e.g. on the originating site, such as 23andMe). While you will not experience any immediate direct benefits from participation, information collected in this study may benefit you and others in the future by helping to determine genetic factors associated with PANS/PANDAS.

If you have any questions regarding the survey or this research project in general, please contact the principal investigator, Bob Horvath, at bobhorvath@alumni.uwaterloo.ca

By clicking on the link below to the upload site, or sending data to the email address above, you are indicating your consent to participate in this study. If you want to contribute anonymously, submit only a fake name and email address at the link. If you use a fake name, make it unique (unidentifiable by others) and make a record of it, in case there is any need to try to contact you (via a comment to this poll in the online groups it is listed in).

https://www.dropbox.com/request/zrejlPbUjAnvN0OuK4Ny

Edited July 7, 2019 by bobh
added email option in Dec., revised title to better distinguish from the followup replication study

[bobh]

I've gotten a comment in another place where this is posted, that some might balk at the dropbox technology of the link in the post, and prefer to email data instead. By all means! My email address is also in the post.

 

I am not expert in security, but for what it's worth, I believe that email is less secure than dropbox (i.e. it is easier for someone else to grab your data). To pretty much eliminate that problem, you could send it anonymously, which by email would mean creating or using an email address that can't be traced to you.

[bobh]

If anyone is still interested in participating, there is only one more week until the cutoff for this pilot study.  I am not sure if anyone from this group has uploaded data (it would be some work for me to check where non-anonymous donors are from, and of course I won't know for anonymous donors), but there are 40 participants so far from other groups.  It is exciting work for which there has not been (to the best of my knowledge) anything published to date.

We intend to publish publicly only at Open Science Foundation, not (at least at this time) formally in any peer-reviewed journal.

[dasu]

Thanks for sharing!

 

[bobh]

Note: This has been approved for posting by Forum administration.

For any that contributed to this genetic study that we posted here (and elsewhere) - thank you so much!  The cutoff was last night, and we had a great result - 71 sets of data.

I will post some tidbits of "aggregate" results here (and in the other groups this was put in).  I am wary of posting full results on facebook, because then facebook could claim some ownership.  Instead, full analysis results (data in aggregate form only) will eventually be posted here:

https://osf.io/pf7q2/?view_only=ba9efeabb38e4a22adced3b5ba4dc5a5

That link above is live to the public now, and you can see (in the last revision of a registration plan), what are the 78 SNPs that we have looked at.

We are going to look at many more SNPs, though, and do lots more analysis than what is listed there.  The plan was just to lay out in advance what we could declare as significant in this pilot study.

[bobh]

I committed to feed results back as they trickle in, so here is one such result:

 

For MTHFR A1298C (rs1801131), there were 70 sets of data:

- 5 were homo for the risk allele (i.e., +/+)

- 23 were heterzygous (+/-)

- 42 of them did not have the risk allele at all (-/-)

 

For MTHFR C677T (rs1801133), there were 71 sets of data:

- 5 were homo for the risk allele (i.e., +/+)

- 26 were heterzygous (+/-)

- 40 of them did not have the risk allele at all (-/-)

 

The first result above is a lower count of risk allele than the general population, and the 2nd is a little higher - but neither statistically significantly so, according to how we setup this analysis. Particularly together, the two of them are pretty much like the general population in terms of number of the risk alleles.

 

I hope this result is not upsetting for any that feel these MTHFR SNPs are a significant player in PANS. Because it doesn't correlate wtih PANS doesn't mean that it isn't something of significance for your particular child. So far, we have only tabulated results like the above (how individual SNPs for these PANS kids fare against the general population). We are still wanting to look at other patterns of many SNPs together among the participants, which might show something statistically significant even if an individual SNP doesn't. It can also be the case that for a given individual, a certain combination of SNPs, including these MTHFR ones, work in a way together to affect the disorder. It is really hard to do either of those kinds of analysis with our genetic data though, so we'll see how our investigations of that works out.

[bobh]

After checking 10,417 SNPs in the PANS genetics data sets that were donated for the above study (from this and other groups), we finally found one that does have a "statistically significant difference" in incidence between 68 PANS kids (who are mostly of European ethnicity), and a representative general population.

This is not at all like Huntington's, where if you have the mutation, you've got the disease, and if you don't, you are clear, 100%. Rather, it is like what we know so far of many immune disorders, where there are many gene mutations (even hundreds of them) that are associated with higher incidence of a disease/disorder (i.e. slightly greater chance of having it with a mutation, but nothing guaranteed either way). For our SNP, almost half of PANS kids don't actually have the risk allele. But a much smaller fraction doesn't have it in the general population; that is the reason it was caught as statistically significant.

So, the interesting thing about this mutation (a SNP allele) is that it has been linked to MS in another formal, published study. When we couldn't finding anything significant with all the "popular" SNPs that people were paying attention to (like MTHFR and FUT2), we eventually branched out to checking ones associated with diseases, particularly autoimmune diseases.

We aren't going to announce the specific SNP just yet, because we are interested in corroborating it with a second set of data. We don't absolutely need that corroboration (because the result is clearly statistically signficant, p=0.00000015 for those that are familiar with p-values), but a replication would make the find so much stronger.

[wisdom_seeker]

Bob, are you still looking for add'l PANS datasets to include in a validation set?   

And as far as the p-value above, did that include some multiple-comparisons correction? (I know that's a not an obvious process, it's easy to over-correct if you do it blindly).

[bobh]

Sorry I didn't notice your response until now.  Well, we are wanting to replicate for good measure, but as a first approach, we are trying to partner with folks that have existing data sets.  They don't know the SNP, and I don't see their data in advance, so I feel that blindness protects us from a certain kind of p-hacking criticism for already-collected data.  If working with others falls down, I might consider doing the replication with data collected as we did in this first round.  I would be happy to let you know if you wanted to contribute.

As to your 2nd question, the simple bonferroni correction was applied for our finding.  So, the target p for significance was 0.05/10,340/2 = 0.0000024 (the /2 was because we were checking both tails), and because we were well under that simple correction, we didn't do anything more sophisticated.   But this target also left some possibilitiies that might be significant as undiscovered.  I have understood bonferroni to be pretty close to the reality - do you know differently?

[bobh]

A report on the pilot study of this post has (finally) been uploaded at Open Science: https://osf.io/pf7q2 . The new files are the last one (the report), and the first one (the spreadsheet).

I am wanting to replicate the finding, and am working towards that.

There is no earth shattering new results there that weren't posted above. For example, the SNP that is statistically significant is still not revealed (because doing so could bias the replication study). The report is several pages of methodology details (the whole thing is 5 pages), so, I don’t recommend people read it unless they are really interested in the how of it all. It’s just that we said we would analyze and report, and so here it is.