Real_S

DNA is not fragile, and can act as a very stable long-term storage medium.

DNA in living cells has some fragility because it is in an aqueous solution and also actively used to generate RNA. DNA out side of cells is even more fragile, it will inevitably be eaten by bacteria. But put DNA in the right sterile environment, it can last for thousands of years and has great resistance to electromagnetic interference.

Also consider that cryptography can be applied directly to DNA molecules.

Cryptography for genetic material:

https://www.biorxiv.org/content/early/2017/06/30/157685

Also, consider that read-speed is rapidly improving. So for long-term storage, reading the data at a reasonable rate in the future is almost guaranteed. Hopefully write-speed will reach a similar rate of improvement.

Importantly, expected corruption of the data stored in DNA is less than what one expects using other forms of data storage. So if stability of the data is a concern, DNA has unique advantages.

Interesting thoughts. Encrypting dna molecules is possible, but it probably wouldn't use a virus. Also, if your cell's DNA was changed like this, it would kill you very quickly, interfering with normal cellular function.

>New techniques ... may soon make the anonymity of their customers’ DNA impossible to safeguard

I strongly disagree.

>He suggested it might be wise to encrypt genetic data to protect personal information.

Maybe we can encrypt DNA molecules themselves! Check out GeneInfoSec

I often find interesting research and/or tech advances outside of my immediate field of study. Also, articles on startups here may have influenced my journey as a founder ;)

And this affects life insurance, so you wont be alive when they dispute your claim!

Also, before you send your DNA off... Keep in mind that these companies are not legally allowed to delete your data.

https://www.bloomberg.com/news/articles/2018-06-15/deleting-...

The intermixing was complex. Many groups migrating and interbreeding at various times.

https://www.nytimes.com/2015/02/20/science/a-new-theory-on-h...

Unfortunately, there is limited interest in genome privacy. There is also a BIG interest in getting people to surrender their genomes.

""But its for the good of humanity, so don't worry about any possible downside.""

Personally, I am trying to develop molecular cryptography, so that genetic data can be protected even if, theoretically, we cannot trust our computers.

>if you want to get the kit and test your DNA out of fun, at least make sure your identity is not obvious.

DNA can be used to identify the person it came from!!!

If you send off your DNA with a different name, there are many ways that they can use to determine that a different name was used. Traits estimated by the DNA could, at least in theory, be used to figure out who you might be. And if your relative is already in the system, your identity may be trivial to uncover.

Short answer is yes, but it depends on the relationship and the disease. For example, if a parent takes the test and is positive for a breast cancer risk genetic variant, then all offspring will have at least a 50% chance of the same result. If, instead of a parent, a grandparent is tested, that percentage drops to 25%.

Unfortunately, GWAS data is unlikely to unlock the secrets of human genetics. We have been in the "post-GWAS era" of genetics for at least a few years now.

> And it's not that big a deal. We will be much better off for it.

I hope you are right! However, I have doubts. Also, just because sensitive information can be compromised doesn't mean that it should be offered freely.

This is an incredibly important point!

Openly sharing your genetic information makes both you AND your relatives more vulnerable, with vulnerability decreasing as relatedness decreases. It also reveals potentially sensitive information about family traits.

These data can be used to identify their source. This will become easier as more data becomes available.

https://news.ycombinator.com/item?id=15209306

I guess the value of the data depends on who measures its worth. Replicated p-values lead to multiple publications, which are valuable to academics. But these results can still cause misdirection. What is a low p-value worth if it doesn't lead anywhere?

This article fails to mention any privacy risks to the volunteers.

GWAS have proven to be, at best, of little value. In fact, due to the unfortunate way that many researchers chase p-values, these analyses often result in misdirection.

What happens? Genome privacy is exchanged for weak science.

Yes, the assumptions they have made to draw their conclusions are highly suspect, especially considering that we have a limited understanding of the genetics of pigmentation. Looks like they are overstretching. My suspicion is that this individual was at least somewhat lighter-skinned than depicted in their recreation. Have you ever seen someone with blue eyes and very dark skin? They exist, but are rare.

But this is just a hunch. My opinion here is not conclusive because they have not released any methods or data. Why not? From the wiki on Cheddar Man, looks like they may have had DNA contamination issues in the past:

"The 1996 results were not subjected to peer review and it was suggested that the sequence was from contaminating modern DNA."

https://en.wikipedia.org/wiki/Cheddar_Man#DNA_testing

Ancient DNA analyses and quantitative predictions are both very difficult, so their results should be considered suspect until actual methods and data are made available.

There is more on this, to be released here with sufficient interest or to be shared with interested parties...

>It is amazing to think that even the rate of mutations is probably an evolutionary response to the environment.

Now consider that mutation is not a single "rate" across the genome. Instead, mutations are targeted to specific regions! Some genome sequences have mutation rates of 10^-3, compared to 10^-9 for most of the genome.

So, when considering mutation in relation to evolution, examine the genes with high mutation rates, especially in their regulatory regions. If fast-changing environmental FACTORS can influence evolution, then hypermutability may be advantageous in genes that interact with these factors. By targeting mutations to specific genomic regions, organisms can "bet-hedge" with specific traits, increasing the probability of survival during extreme environmental shifts.

Think about growth rates in relation to boom-bust food cycles, or neural development in relation to dynamic social conditions. What evolutionary strategies would be optimal for rapidly changing environmental factors? Increasing mutation rates genome-wide is not the answer, as you point out it leads to problems with viability. Thankfully, mutation rates are local and traits are bet-hedged against variable environments.

What a wonderfully variable species we are! We adapted to different environments worldwide! And it only took tens-of-thousands of years?! Amazing :)

Interesting connection with brain complexity, but note that the function of the Huntinton's gene remains poorly understood.

Something that may be related... The same unstable repeat that causes the disease is ancient and is conserved throughout mammals! So other mammals can get a similar disease driven by the same mutations. This is useful for generating models of the disease in other mammals (e.g. mice). Such conservation of repeats is exceptionally rare.

There is something important about the type of mutations that drive Huntington's.

Huntington's is caused by an expanded repeat composed of (CAG)n with n>40 repeats. Due to the instability of repeat regions, new mutations arise in this gene at a relatively high rate. So, even if there were strong selection against expanded repeats the disease would continue to exist. However, as mentioned by others here, strong selection is unlikely due to the late age of onset of the disease. So mutation continues to introduce the disease and selection has been too weak to remove it.

In contrast, BRCA mutations differ from the Huntington mutations. BRCA mutations are single "point" changes in the sequence, known as SNPs, that disrupt the protein function. Novel point mutations are exceptionally rare.

So, unlike with BRCA, when investigating why Huntinton's disease exists, one must consider BOTH mutation and selection.

1. You are perceptive, this is usually something to watch out for. However, it does not always indicate that a study is flawed, just that the research might have met with unwanted resistance, and the authors are in position to avoid that resistance.

2. Yes, these scientists are looking to make their research popular, highlighting an unfortunate conflict of interest present in academia. They do this to get big grants. Important observation, but insufficient to dismiss a finding out of hand.

3. So many authors, such a joke! These people mostly just help collect data, or sometimes simply share data they already have and then give the manuscript a quick read before publishing. But that is actually beside the point. If a lab already had data ready, and isn't pressured to add authors for political reasons, then two authors is not necessarily a red-flag. In fact, these analyses are relatively easy to do with available software. If they have the data, they just plug and chug.

4. The findings of the paper suggest strongly that friends are more genetically similar, on average, akin to fourth cousins. The results have yet to be rebutted to my knowledge. The fact they found some regions were negatively correlated does not conflict with their overall results. I have some doubts that their regional analyses should carry much weight (e.g. olfactory and immune regions). Furthermore, their explanations using natural selection appear to be a stretch. These are possible overstatements of results, allowed because of your points 1 and 2.

My overall impression of the paper is that main results are somewhat interesting and entirely novel. But there is a ton of stretching for just-so stories and "interesting" findings that will be unlikely to hold up. Typical PNAS.

Population structure is controlled through the use of the structure as a co-factor in the analyses. They use a principal component analysis on the genotypes to estimate the structure, and then use the first few principal components (PCs) as cofactors in the analyses.

So if the European population they used consisted of half English and half French participants, then the first PC would likely divide the two. So if British or French ethnicity affected the analysis, this effect would be reduced or eliminated using just the first PC.

How many PCs should be used? Well, ten is a common number chosen. Is it enough? Probably for most studies including this one.

So ancestry and ethnicity in this type of research emerges purely from a PC analysis on the genotypes. They do not use self-reported ethnicity or any other type of common ethnic grouping. To control for population structure, self-reported ethnicity is rarely useful. Its far too unreliable. But these groupings do sometimes loosely correspond to the PCs found in the analysis, supporting the use of PC as a method to control for ethnicity.

Interesting conflict between these researchers.

From Erlich [0]:

"The take home message should be that identifying someone in a group of ten people requires very little effort. Anyone with access to even low dimensional data, such as basic demographic, can do that. This is not very surprising."

"To summarize this point, the title says: “Identification of individuals by trait prediction using whole- genome sequencing data” but most of the trait predictions is carried by ethnicity of the individual (genomic PCs) rather than the trait specific SNPs."

Summarizing: DNA can estimate ethnicity, which can be used to predict a range of traits, including facial structure. These results would be far more impressive if they were able to predict faces in data composed of a single ethnicity. HLI may have overstated their results, but nevertheless they raise important points about re-id.

[0] http://www.biorxiv.org/content/early/2017/09/07/185330.1

After looking through the Contents of both:

CASI looks very cutting edge, but also covers the origins of the use of computers for statistical methods. So this covers how computers can and have been used to do statistics.

FoDS looks like a very thorough data science book. Less of a "then VS now" book and more of a collection of what you need as a data scientist.

In CASI, I like the chapter on FDR, a very important topic not found in FoDS (?!). FDR is critical for correcting for multiple testing, seems essential for data science but maybe the authors consider it cutting edge and not foundational. However, the wavelet chapter in FoDS makes me happy, a very useful topic for series data.

Both great books, thanks for the links!

If you follow their methods, specifically [63], they are modelling the proportion of null hypotheses in their tests. Performing FDR by group is a reasonable approach, and does not necessarily cause P-value inflation. This is different from a Bonferroni correction, in which total numbers of tests done is the only factor. These FDR models of the proportion of null hypotheses rely on the distribution of p-values, not their raw count (assuming a sufficient count of P-values is present in which to model). Dividing the FDR correction as they did provides more information about group-specific influences, and should not be considered p-value hacking.

Nevertheless, thank you for calling this out. There is way too much p-hacking in science, especially biology.

[63] Benjamini Y, Krieger AM, Yekutieli D (2006) Adaptive linear step-up procedures that control the false discovery rate. Biometrika 93:491–507.

Thank you for clearly sharing how and why scientists can accurately measure genetic effects in twin studies. You are correct in your assertion that mono vs di - zygotic twin studies are usually good measures of heritability.

I would add a slight technical caveat. Twin studies are excellent at measuring additive genetic effects, i.e. when different genes add together nicely to contribute to a trait (height is a good example). Twin studies typically assume that genes do not interact in funny ways (e.g. epistasis). Thankfully, for many, but not all, traits examined the additive genetic effect catches most of the heritability.

If epistatic interactions drive a trait, however, the statistical models typically used in twin studies would actually under-estimate the influence of genetics.

Further reading: http://www.apa.org/monitor/apr04/second.aspx

The breakdown of ethyl alcohol is as follows:

alcohol dehydrogenase ethyl alcohol -> acetaldehyde

aldehyde dehydrogenase acetaldehyde -> acetate

Much like formaldehyde, acetaldehyde is a toxic mutagen. Depending on your diet, genetics and the health of your liver, alcohol consumption can lead to an accumulation of acetaldehyde. This is why many asians get "red-faced" from drinking, they actually accumulate acetaldehyde due to their genetics. The other major toxin in many alcohols is methyl alcohol, which can break down into formaldehyde, but is actually pretty toxic on its own.

These are fun facts to share at drinking parties ;)

This is an awesome wealth of videos! Check out the Data Visualization section for some Tufte inspired lectures.

These guys really know their bullsh!t :)

No two cells are ever genetically identical. Within a common hyper-mutable region, tandem repeats, the mutation rate is 10^-3 to 10^-5. There are over 10^5 tandem repeats in the genome, and therefore at least one mutation is expected for every cell division. Many other types of hyper-mutable regions exist in the human genome.

In this research they only examine one form of genetic variation, SNPs. These findings only reflect a small proportion of the somatic variation present in the body.

There is no real surprise in these results, but the data may nevertheless be useful!