Polygenic info used to identify risk of AD

[NF52]

I really wonder how they can have a 100 percentile unless they have the deterministic PSEN and APP genes in there?

I think the PHS is a relative metric, not an indicator of absolute risk. "100" means maximum risk relative to other humans, not 100% risk.

Good point, Marc.
Here's a quote from the Dash sample report:

A PHS percentile of 99 implies that 99% of people are estimated to have an equal or lower risk than such individual. An individual with a high PHS percentile is estimated to have a higher risk than the one with a low PHS percentile, at any given age.

At age 66, with ApoE 4/4 and family history, I am sure I have a higher relative, statistical lifetime risk than my 66 year old ApoE 3/3 husband, or almost anyone else who is not ApoE 4/4 and my age. After all, jkramer65 and I are in the rare 2% of the population, so it make sense that I might be in the top 2% or so of risk. But statistically, that does not mean I have a 98%-100% chance of MCI or dementia by age 85, I believe. More like 50% probably, given my family history and the general population incidence of AD after age 85. Hopefully lifestyle and other protective and preventive measures move my actual risk either way down, or pushes it further back.
As a comparison, I probably also have a higher risk of a sunburn that 98% of the population given my skin tone (and previous basal cell skin cancer on my nose). But that does not translate to a 98% chance of malignant melanoma or even severe sunburn if I manage the environmental factors better now that I have that information.

[circular]

Thank you both for the clarifications

[jkramer65]

My PSEN and APP are all green on 23 and me so not sure how that impacted my score. I guess 50% at age 80 is better than age 60??? I failed to mention I also have a 69yo sister who is mentally intact. She hasn't had any genetic testing. Who knows right? The bottom line is we have no clue I guess.

[jkramer65]

I really wonder how they can have a 100 percentile unless they have the deterministic PSEN and APP genes in there?

I think the PHS is a relative metric, not an indicator of absolute risk. "100" means maximum risk relative to other humans, not 100% risk.

Good point, Marc.
Here's a quote from the Dash sample report:

A PHS percentile of 99 implies that 99% of people are estimated to have an equal or lower risk than such individual. An individual with a high PHS percentile is estimated to have a higher risk than the one with a low PHS percentile, at any given age.

At age 66, with ApoE 4/4 and family history, I am sure I have a higher relative, statistical lifetime risk than my 66 year old ApoE 3/3 husband, or almost anyone else who is not ApoE 4/4 and my age. After all, jkramer65 and I are in the rare 2% of the population, so it make sense that I might be in the top 2% or so of risk. But statistically, that does not mean I have a 98%-100% chance of MCI or dementia by age 85, I believe. More like 50% probably, given my family history and the general population incidence of AD after age 85. Hopefully lifestyle and other protective and preventive measures move my actual risk either way down, or pushes it further back.
As a comparison, I probably also have a higher risk of a sunburn that 98% of the population given my skin tone (and previous basal cell skin cancer on my nose). But that does not translate to a 98% chance of malignant melanoma or even severe sunburn if I manage the environmental factors better now that I have that information.

I support this theory NF52! 50% by 80 isn't the worst I guess. Makes sense with my family history actually. It is just harsh to see 100% for the polygenic score cs even 99. Really? NOONE has a higher risk than me? Ugh. I have to think one person on earth must! haha.

[circular]

It is just harsh to see 100% for the polygenic score cs even 99. Really? NOONE has a higher risk than me? Ugh. I have to think one person on earth must! haha.

This probably isn't much comfort, but I'm guessing you at least have company but just don't know who they are yet.

I wish the report would return our top five good and top five bad so we'd have some things to celebrate and other to try to work on if they're the least bit understandable or actionable.

[Fiver]

Never tell me the odds - Han Solo.

[NF52]

It is just harsh to see 100% for the polygenic score cs even 99. Really? NOONE has a higher risk than me? Ugh. I have to think one person on earth must! haha.

This probably isn't much comfort, but I'm guessing you at least have company but just don't know who they are yet.

I guarantee at least one person on earth has a higher risk than you do--anyone who already meets the criteria but has not yet received a diagnosis of AD or other dementia. On a more practical basis, remember that a percentile risk of 99 or 100 is based on a group of 100. Not speaking as a statistician here, but just saying that when we say someone has a 100% "risk" of being the tallest person in in an average group of 100 people, that does not mean she/he is the tallest person in a group of 1000, or 5000 or in a group of NBA players.

The DASh sample report says that the model looks at:

similarities between your genetic profile and the average risk in the group analyzed in the research.

By definition, the "average" risk of Alzheimer's is skewed to much lower than the risk of the ApoE 4.Info forum all hanging out in a stadium together on a nice fall day. So in an average group of 100, we are the "tallest" kids in the room. In the stadium, we look more alike than different and the PRS may well fail to differentiate between us on a multitude of factors.

I'm with "Fiver"--don't tell me the odds-- at least until you've studied that large healthy group in the stadium over a long period of time.

[Fiver]

I do appreciate the scoring system. I have learned that the process of making a scoring system and improving it is an excellent way to learn what we know and what we don't know, and move science along. But I don't think we know enough to create an accurate scoring system for an individual.

....also I need to save my $$ for supplements and avocados (which seem to somehow go from unripe to over-ripe as soon as I look away). Personally, I just assume I am in the flashing red zone and keep going. But I can understand how it might be helpful and motivational for others.

[jkramer65]

I also went through every SNP that was on my promethese report and only a few of them on the PHS scale were of concern. Basically we don't know ANYTHING! If it makes me live a better life then it is worth it right? I have one sister diagnosed at 63 and another fine at 69. My 92yo mother has all her marbles and my Dad died of AZ at 87. It is a crap shoot. Being so much younger than everyone gives me some hope that medical science will have more info as I grow older. Meanwhile I'm exercising every day, doing my morning brain exercises, eating well -- which I always have -- and taking supplements. AND trying not to be obsessed and continually freaking out that I'm at a higher risk. I need to enjoy life and bank as much money as I can for possible caretaking. I just have bad days some times. I'm sure everyone does.

[xingxu]

@MarcR

Thanks for sharing the old version 23andMe download data with us.

We ran the old version data with our algorithm, we got the identical results as the new version of data in your case.

We took a close look at the difference between the two files,
We saw
565 unique additions from the old download (2013)
3002 unique additions from the new download (2018)
340 content changes between the two files.

Because the algorithm only uses loci that are located in autosomes, commonly shared by 23andMe and ancestry and SNPs only, loci in chrX, chrY, and chrM, or starting with i (23andme specific loci), or with indel genotypes in the unique additions, should all be removed.
88 unique additions in the old download (2013) are possibly qualified loci for the analysis.
61 unique additions in the new download (2018) are possibly qualified loci for the analysis.
14 content changes between the two files.

Finally, comparing these possibly qualified rsids to our list of high-confident loci (about 600,000),
there are only 3 overlaps from unique additions of the old download (2013) and 0 overlaps from unique additions of the new download.

Thus, we don't believe minor changes between different versions of 23andme data should affect the results significantly.

I don't think downloading the same data from 2013 would give a different result. Did you test again using the more recent version and then use that data?

I have only tested once, in 2013. The Change Log subarea of the
Raw Data Download area clarifies. For example, here's the entry pertaining to the 7/27/17 update:

As part of our continuous efforts to improve the quality of data present in your raw data download, the number of SNPs available in your download may have changed.

I used the Unix diff program to compare the 2013 and 2018 downloads and confirm that the 2018 download contains more information.