Polygenic info used to identify risk of AD

[J11]

The Supplemental Figure (from the supplement of the PHS article) that I have uploaded above clearly reveals the significant shifting of the social/political landscape that will ensue as this research is more fully appreciated by the wider community.

There is a mainstream awareness of the APOE epsilon 4 as a risk enhancer for Alzheimer's. It should now follow that there will in time also develop a broad awareness of the PHS scale (possibly even the epsilon 4 designation will be subsumed by the PHS as a definer of AD risk). As can be seen above there is a certain amount of overlap between epsilon 33s and epsilon 34s, and between epsilon 34s and epsilon 44s. For many 34s, learning that they had an AD equaling or exceeding that of some 44s would obviously be unwelcome news. Of more social impact will be the change of perception that occurs in those high percentile epsilon 33s when they realize that they might have risk near to that of 34s.

I am aware of several families who had no pre-warning that they were at risk for dementia. If PHS will be able to prospectively provide this information (as the article suggests could be possible with additional research), then such families will be able to better prepare for managing the illness.

[Brugman4]

Bergman4, thank you for the suggestion.

It is a great mystery to me about the dbsnp track.
It is showing in my Plot Tree. So, when I launch the program it is a track, though for some odd reason the
information for dbsnp 147 is not available to the program to convert the 23andme file.

I checked my Local folder and it is not there.
From what I can see it is not in any of the folders.
Perhaps it is an external file (that wouldn't be sensible).

I suppose they want you to download the newest and greatest dbsnp 149 ( the full version of 2.9 Giga no less!)

By the way, might it be possible to reconstruct your steps to getting the Alzheimer Annotation PHS file up and running and post this information to the thread? Others might also be interested in knowing exactly how to upload it to GenomeBrowse.

You might have noticed in past posts that there was a slight trouble with setting up the local folders to accomplish this.
I would provide this description myself, though it seems that after you done the initial setup properly that is water under the bridge.

I think it would be great to have another one of these Annotation files, perhaps this one could have all the PHS SNPs, along with the rest of the IGAP SNPs and a few others that have appeared in the research.

My dbsnp is in User Annotations. I have the same problem with my Genos annotation, it displays as a track (a 2nd related track is showing only questions marks tho) but it's not in any folder.

I'll give the PHS step-by-step a shot.

It would be great, too to have versioning on the PHS SNP Annotation file, so as research comes out, we add in the latest SNPs and then post it so folks can upload the latest version.

Progress on 23andme: I was able to upload my 23andme file but no success yet with plotting to get the track to display. It needs a companion annotation, so (1) download Reference Sequence GRCh37 g1k, 1000Genomes, (2) refresh, (3) add/convert your 23andme file.

[J11]

The PHS research has (at least) for me changed the dementia conversation.

I will watch with interest as this new insight permeates through our community.
It took some time for our forum be forged with purpose and a voice to advocate and advance the interests of those with epsilon 4. How will the 50 million epsilon 33 Americans who are in the top quartile of PHS dementia risk amongst epsilon 33s respond when they become aware of this risk? I would think that they would eventually create a forum similar to our own. Perhaps we should help them find their sea legs by budding off a sub-forum on our forum. Without doing so might mean that they would be adrift for a few years until they
recognized that they were all in the same boat (so to speak).

As the PHS is a derived score from SNPs some of which need to be imputed, widespread recognition of its importance to those individuals at risk could take some time. Yet when snpedia reports the PHS scores, such recognition would be greatly accelerated.

[J11]

Bergman4,

that's the weird part: my dbsnp file is no where! i checked User Annotations!
I mean doesn't it have to be somewhere if it is a track in the program? Or can you just have phantom tracks that just pop out of intradimensional space?

With the Genos file, I think I could just drag and drop it into the program window and it just plotted by itself.

My first guess would be that the PHS annotation will now be fairly stable through time. They included nearly all of the significant IGAP regions and even fine tuned a few regions into opposing SNPs betas. Possibly from here on, the substantive change in the PHS will occur by adding additional AD genes.

I would love to have a better understanding of what the researchers are claiming regarding how to interpret their PHS. Much of the commentary language spoke in terms of heritability and area under the curve which is not applicable to the PHS. As one of the authors noted it is more of characterizing people into risk bins related to their disease risk. I'll have to reread the article to see how accurately they could do this binning.

One of the exciting aspects of the research is that it has provided easily verifiable assertions. With epsilon 4, after the initial research was published, it was accessible for any other group to genotype 1000 AD patients/controls to confirm the finding. A single epsilon 4 increases risk of dementia 2-3 times.
This new research found that the top quartile of 33s in the PHS had a 3 times increased risk versus the bottom quartile. This claim should likewise be readily confirmable. (Too bad that the scientific publishing model takes years to get to press. There must be standing AD genotype sets that could instantly report confirmation or non-confirmations.)

Have you been able to actually get the 23andme file as a track on GenomeBrowse? I thought all that was needed was the dbsnp file?

[J11]

Here are the numbers for the PHS.

If anyone can figure out how to move the number in the bottom right corner the Total Total beta to -0.05, I would be very interested to know how this was done.

There are still a few confusions with the numbers.

One confusion relates to the choice of the risk allele.
Of the 31 genotypes in the PHS (aside from epsilon) only 4 are in the major alleles (4,11,16, 29). This makes a great deal of sense: mutations involved in brain function cause dysfunction more often than they result in benefit. Column I modifies the "MAF" to reflect the risk allele present.

Another modification is when the betas are negative. In this instance, the major allele has been substituted for the minor allele. The idea is that the calculation one expects to occur is the number of major alleles is multiplied by the beta with a negative beta.

I am not sure whether the formulae carry over when the spreadsheet is uploaded, though if you click on a number on the top of a column, then an expression should pull up in the formula line of LibreOffice or other program. You can then see the formula for the cell if present ( the derived columns such as J,K,L,M) should all have formulae. To make adjustments you can reenter new formuale. Do this by clicking on a cell, in the line to the top, erase what is in the line, type =, and then e.g. Product, Sum etc, click on cells to include and separate elements with a comma, start and end the formula with an opening "(" after the funciton word and a ")" at the end. If you like the formula in one cell, you can extend the same formula for the entire column by simply dragging the right bottom of the initial calculation cell where there is a small black rectangle.

The Total Beta in Column O is the expected beta for each SNP. One could place their genotype betas beside this line to see how they compare with the average.

PHS calc.ods

[J11]

I am quite impressed by how many of the PHS SNPs can be found in the Michigan Imputation results.
I have found almost all of them there.
The authors might have even used this service to conduct their research.
If so, I wonder why they would not have included the full 40 million imputations?

LG1, might you ask the author of the PHS article exactly how the calculation for the PHS score was done?
I have all the genotypes I need for the PHS, though the numbers are not "adding up".

These polygenic scores are typically simple sums.

My best guess as of now is that the positive betas should be added once for one risk allele and twice for two alleles.
The negative betas probably should be added once for a non-risk allele and twice for two alleles.

As a specific example [positive beta], those with the PICALM SNP rs543293 (where A is the risk allele) would add:
0.60 if they were AA (two risk alleles)
0.30 if they were AG (1 risk allele)
0 if they were GG (0 risk alleles)
to their PHS,

and [negative beta] those with NME8 SNP rs17265593 (where T is the risk allele) would add
-0.46 if they were CC genotype (2 non-risk alleles)
-0.23 if they were CT genotype (1 non-risk allele)
0 if they were TT genotype (0 non-risk alleles).

A confirmation of this interpretation would be very helpful.

[Harrison]

75% of AD patients do not have an epsilon 4!

I do want to ask about this comment that 75% of AD patients do not have an episilon 4. Since the use of amyloid imaging has become more common, nearly every study I have seen quantifying the percentage of patients with an episilon 4 puts it at around 60%. That is a pretty big discrepancy!

BTW, I checked my v4 23andme raw data against the spreadsheet, and found 20 SNPs were not found at all, 5 SNPs were "not genotyped" and only 8 SNPs were genotyped.

[J11]

Um, like OK wow, sometimes there just aren't words.

This just in:

" The U.S. Food and Drug Administration today allowed marketing of 23andMe Personal Genome Service Genetic Health Risk (GHR) tests for 10 diseases or conditions.

...
risk for developing any one of the following 10 diseases or conditions:

...
Parkinson’s disease, a nervous system disorder impacting movement;
Late-onset Alzheimer’s disease, a progressive brain disorder that destroys memory and thinking skills;"

viewtopic.php?f=16&t=3072

Am I the only one wondering whether this means the PHS?
Could this already be launching?

Sort of what we have been talking about on this thread.
The PHS will have a revolutionary effect on the way many in the community perceive their AD risk.
Appears that the FDA wants to get on the right side of the tidal wave and let genetic technology be a
tool for the good.


I have seen a fair number of people wake up one day and out of the blue, pretty much from outer space they find that a loved has dementia. They had no ability at all of how to cope with that. Total catastrophe. The PHS will allow people to more effectively manage their dementia risk. It is not perfect now, though I have no idea why people would not be scrambling right now to find out their PHS.

The high percenters would be well advised to at least start think about how they might cope.

From the PHS article:

"Within the ADGC Phase 1 cohort, individuals in the highest PHS quartile developed AD at a considerably lower age and had the highest yearly AD incidence rate. Among APOE ε3/3 individuals, the PHS modified expected age of AD onset by more than 10 y between the lowest and highest deciles (hazard ratio 3.34, 95% CI 2.62–4.24, p = 1.0 × 10−22). In independent cohorts, the PHS strongly predicted empirical age of AD onset (ADGC Phase 2, r = 0.90, p = 1.1 × 10−26) and longitudinal progression from normal aging to AD"

I think I called it the highest quartile as 3.34, above quote notes that it is the highest decile.


"Among APOE ε3/3 individuals, who constitute 70%–75% of all individuals diagnosed with late-onset AD, age of onset varies by more than 10 y, depending on polygenic risk profile (Fig 2). At 60% AD risk, APOE ε3/3 individuals in the first decile of the PHS have an expected age of onset of 85 y, whereas for individuals in the tenth decile of the PHS, the expected age of onset is greater than 95 y. These findings are directly relevant to the general population, where APOE ε4 accounts for only a fraction of AD risk [3], and are consistent with prior work [26] indicating that AD is a polygenic disease where non-APOE genetic variants contribute significantly to disease etiology. "

There might be different definitions of AD used. If you move into an increasingly geriatric population then perhaps everyone will have some form of dementia. The article might be leaning more toward this definition. However, in a more typical clinical context for example in a trial of people 60-80 years of age you might see more of a 50-50 split. Suppose you could use the PHS model to calculate the numbers.

[J11]

Thanks Harrison for the support.

I am thinking that some of us can do the leg work and then the next wave will be able to reach the shore a little bit easier.
As you have probably read there have been a few rough patches with GenomeBrowse, imputation, understanding the PHS etc. though we have made a fair amount of progress. This should make it much easier for the second wavers.

With the Michigan Imputation Server, looks like a super great service. It can take a few days to get your results back, though their mega high volume imputers so they have a very firm grasp of what they are doing, I have just been double checking my numbers and the Imputation appears to have now filled in my entire PHS. If someone with only a 23andme file might try the server it would be interesting to see whether they would also have all the genotypes ( my exome file might have filled in any gaps).

The only disappointment that I have now is that 4 of the imputed SNPs have only OKish quality scores. 3 of them are least 90% or better probability while one of them is only 75%. At 75% it is almost down to a coin toss. The sequencers like to have probabilities of being incorrect with a few leading zeroes. Fortunately all four of the imputed gentypes are reporting the largest PHS, so any error would reduce my AD risk. However, if the PHS were 0.3 less than I have calculated I would want to know. PHS gene chip? Perhaps 23andme or someone could offer it.

[Harrison]

"Among APOE ε3/3 individuals, who constitute 70%–75% of all individuals diagnosed with late-onset AD, age of onset varies by more than 10 y, depending on polygenic risk profile (Fig 2).

There might be different definitions of AD used. If you move into an increasingly geriatric population then perhaps everyone will have some form of dementia. The article might be leaning more toward this definition. However, in a more typical clinical context for example in a trial of people 60-80 years of age you might see more of a 50-50 split. Suppose you could use the PHS model to calculate the numbers.

Thanks for clarifying. They don't say where the 70-75% number came from, so I am assuming it came from the ADGC GWAS data. I don't know anything about that collection, but it's possible they don't have amyloid measures and are just looking at all-cause dementia. I do know that other datasets, such as ADNI, which do have some amyloid markers, match up with the 60% E4+ number that most pharmaceutical clinical trials with Amyloid imaging report. I think it's quite possible that E3 and E4 dementia cases could be quite different, and the polygenic risk score might help clarify this.

I will look into the Michigan service and see if it can fill in the gaps in the 23andme SNPs.