Reconstructing a generation unsampled

By Razib Khan | January 14, 2012 3:59 pm

In the near future I will be analyzing the genotype of an individual where all four grandparents have been typed. But this got me thinking about my own situation: is there a way I could “reconstruct” my own grandparents? None of them are living. The easiest way to type them would be to obtain tissue samples from hospitals. This is not totally implausible, though in this case these would be Bangladeshi hospitals, so they might not have saved samples or even have a good record of hem. Another way would be to extract DNA from the burial site. This is not necessarily palatable. But assuming you did this, if you have access to a forensic lab it might be pretty easy (though I think most forensic labs using VNTRs, rather than SNP chips, so I don’t know if they’d touch every chromosome), I’m not sure that the quality would be optimal for more vanilla typing operations, especially for older samples which are likely to be contaminated with a lot of bacteria.

For me the simplest option is to look at relatives. Each of my grandparents happens to have had siblings, so there are many sets of relatives related to just each of those individuals of interest. I also have many cousins, so pooling all the genotypes together and using the information of a pedigree one could ascertain which chromosomal segments are likely to derive from a particular grandparent. To give a concrete example, my mother has a maternal cousin to whom she is quite close. By typing my mother and her cousin one could infer that the segments shared across the two individuals derive from the common maternal grandparents. Of course there’s a problem that cousins have a coefficient of relatedness of only 1/8th, so there is going to be a lot of information missing. But, if you had lots of cousins you could presumably reconstruct the genotypes far better.


But what if you didn’t have any of this? I came up with a crazy idea, and I want to throw it out there to see how crazy it is. The issue from the perspective of you, the indivdual without grandparental information, is that for either your mother or your father you don’t know which homologous chromosomes come from which parent (your grandparents, their parents). As it happens, everyone has a male parent and a female parent. So if you can assign a a chromosomal region as having come from the male, and another as having come from the female, then you can reconstruct some of your grandparents’ genotypes because you know their sexes. How can you make this determination?

Genomic imprinting. This is a phenomenon where genes from a given parent, often of a particular sex, are expressed, while those of the other sex are repressed (often it manifests in terms of methylation or lack of methylation). Therefore, if you have a gene, A, which is usually expressed if inherited from a male parent and repressed if it is inherited from a female parent then the state of that gene within a chromosomal region can be a “tag” for the sex of the parent of origin. With enough of these imprinted genes you can create a mosaic of the genome of the individual in terms of sex of origin. Obviously genomic regions from different sexes are from different parents. If you have enough children of these two parents you should be able to infer the whole genomes of these individuals.

The big reason this probably won’t work is that there just aren’t enough imprinted genes in the human genome. But what do readers think?


Comments (6)

  1. I think it sounds a lot easier to just try to get samples from hospitals. 😉

    You could probably do a fair amount of genetic forensics by combining the cousin and imprinting investigations, though it could get financially burdensome if you have a large family (but on the plus side, the bigger the family, the bigger the pool of evidence).

  2. #1, yeah, the $ would get bigger to typing so many people. OTOH, i assume it will decrease A LOT in the near future. and though i didn’t elaborate, probably geographic/regional sampling will get good enough that you can differentiate ancestors even if you aren’t recently mixed-race. e.g., if your maternal grandfather was french and maternal grandfather was a mix of swedish/english/dutch, the analytics would be good enough to ascertain that. with my grandparents, two of them came from regions 100 miles apart. knowing how genes cluster in european villages that’s probably enough distance that they could be differentiated.

  3. SCulliton

    This would require you know MANY genes which are imprinted with a parental bias, and of course the bias rather than random. For example imprinted gene X is always inherited maternally or paternally. I think you hit it on the head when you said we don’t have enough known imprinted genes in the human genome.

    Furthermore, it is known that environmental factors can have an affect on our epigenome, including imprinted genes. We would have to know if a particular gene is methylated via parental bias imprinting or altered methylated via envioronmental affects. For example this study,, gives an example in which imprinting is altered via maternal diet. There may also be other ways in which imprinting is non-permanent.

    You might find this of interest:

  4. This would require you know MANY genes which are imprinted with a parental bias, and of course the bias rather than random. For example imprinted gene X is always inherited maternally or paternally.

    just a note

    1) re: number, do note though that these chromosomes are inherited in large blocks, though with some recombination. so we wouldn’t need to know thousands of genes. just enough to “tag” the several dozen homologous chromosomes.

    2) obviously it would best to have totally distinctive expression…but even moderately biased would probably allow you to create some sort of statistical model. but then you’d need WAY MORE genes.

    but it does seem that this is not workable as of now….

  5. Grey

    It sounds like it could be the basis for an interesting episode of CSI or House. I can imagine some potential with ancient DNA too, maybe.

  6. Roger Bigod

    If a hospitalized patient had surgery and slides were made for a histological exam, the paraffin blocks containing the tissue are saved by US hospitals, and AFAIK worldwide. A few years ago I saw a paper that stated the DNA is essentially intact for 5 years, and after that starts to degrade slowly. There may be technical improvements now.

    Thomas Jefferson’s Y markers were determined from a lock of hair. I don’t see why the whole genome should be a problem.

    As you go further back or try to capture close to 100% of an ancestral genome, the numbers of samples required goes up exponentially.


Discover's Newsletter

Sign up to get the latest science news delivered weekly right to your inbox!

Gene Expression

This blog is about evolution, genetics, genomics and their interstices. Please beware that comments are aggressively moderated. Uncivil or churlish comments will likely get you banned immediately, so make any contribution count!

About Razib Khan

I have degrees in biology and biochemistry, a passion for genetics, history, and philosophy, and shrimp is my favorite food. In relation to nationality I'm a American Northwesterner, in politics I'm a reactionary, and as for religion I have none (I'm an atheist). If you want to know more, see the links at


See More


RSS Razib’s Pinboard

Edifying books

Collapse bottom bar