Parent/Child Genome Comparison: 100,000 Years Ago

[fluidistic]

Let's say you know you are given the full genome of 2 individuals such that they are related as father/son or mother/daughter, that they both died at age 30 some 100 k years ago. Can you tell who is the parent/child by looking at their genomes only? If so, how?

I picked 100 k years only because I want to remove C14 and other isotopes out of the question, i.e. you cannot determine who is the parent by checking via that method. I picked 30 years old so that their telomeres have the same length. No need to insist on the point that after so many years, the genome, if any, won't be full. I don't care about that, it's hypothetical, etc.

If the answer is yes, then I think this implies that evolution has a direction, so I guess that the answer is no. But my gut feeling thinks the answer is yes...

 

[Vanadium 50]

If a parent shares 50% of their genes with their offspring, don't the offspring share 50% of their DNA with a parent?

 

[fluidistic]

If a parent shares 50% of their genes with their offspring, don't the offspring share 50% of their DNA with a parent?

I think that approximately yes, there are cross overs, mutations and I guess many other subtleties. But I don't really see how this helps answering the question. What am I missing?

 

[jedishrfu]

I think you can’t know unless there's a clear age difference (telomere length) or you have dna from other family members and know their relationships.

 

[Vanadium 50]

What am I missing?

That you can only tell if there is some asymmetry in the system.

 

[BillTre]

If the answer is yes, then I think this implies that evolution has a direction, so I guess that the answer is no.

Not sure what you mean here by direction. There are many possible interpretations.
I would say yes evolution has a direction: ancestral to derived (or parent to offspring).
However, there are also other possible evolutionary directions, which could be applied (in a right or wrong way).

Ways to tell offspring from the parent genetically, assuming perfect knowledge of the genomes:

• Location in the genome of identical sequences between the two. Larger sequences identical between the two would indicate the parent whose DNA has not been shuffled around by recombination before being passed on to its offspring. If there were no crossing over, then there would be no differences (except for new mutations in the second generation) between the two genomes. The offspring would inherit whole intact chromosomes from each parent. One chromsome would be entirely Mom and the other entirely Dad. Crossing over would mess this up and the offspring chromosome pair would be a two chromosomes with a mix of sequences from each parent linked together in a single chromosome.
• As @jedishrfu said, the length of the telomers would indicate which is older (the parent), but if both samples were from the same age, this would not work.
• There should be new mutations occurring between the two generations. These could be any of a number of different kinds of changes in the genome. Some of the changes like point mutations would be difficult to tell which way the transition went. However, some like copying and insertion of transposons (or perhaps bits of viruses) would appear as a novel addition to the genome of one when compared with the other. The offspring would be more likely to have the new insertion.
• There may be other ways to do this.

 

[sysprog]

I think that if one of the specimens is allelically homozygous for a given gene, and the other is allelically heterozygous for that gene, it is likely that that the other is the offspring.

 

[Ygggdrasil]

I think that if one of the specimens is allelically homozygous for a given gene, and the other is allelically heterozygous for that gene, it is likely that that the other is the offspring.

I don't think this is true:

AA x Aa --> Aa (homozygote is the parent and heterozygote is the child)
Aa x AA --> AA (heterozygote is the parent, homozygote is the child)

 

[sysprog]

I don't think this is true:

AA x Aa --> Aa (homozygote is the parent and heterozygote is the child)
Aa x AA --> AA (heterozygote is the parent, homozygote is the child)

Clearly both are possible, however, which do you think is more likely?

Assuming only two possible alleles for the gene, the Aa heterozygote has only a 50% chance of contributing allele A, whereas the AA homozygote has a 100% chance of contributing allele A. The other parent can be AA, Aa, or aa, if the heterozygote is the offspring, but if the homozygote is the offspring, neither parent can be aa; the other parent would have to be either AA or Aa $-$ filling out the table, it seems to me, would show more cases in which the AA was the parent.

 

[BillTre]

This depends on the particular alleles involved (and which ones are common (and how common) in the breeding population), as well as the kind of breeding involved ((inbreeding, outbreeding, or other schemes).

The original question was based on a complete knowledge of both genomes. This would allow the alleles at more loci to be surveyed.
A @Ygggdrasil type analysis would have to be applied to the alleles at each loci, giving a more statistical kind of result.

To me, the normal assumption would be that there are several (or even many) loci with different alleles to consider in an outbred population.
I would not consider an inbred population, unless stated. This would be a different situation with few versions(or only one) available at any loci.
In various populations, there can gradations between fully inbred and highly outbred populations.
If there were more that just two alleles/loci, then conclusions would be much more clear.

A fully inbred line (no allelic variation) would negate my first point (above), since all stretches of corresponding sequence would be the same.

The problem, as stated, does not really make clear these details of the population's structure.

 

[Vanadium 50]

I think the problem is if you can do this in general. Sometimes of course you can get lucky. If you have a mutation that you know induces sterility, you know which is the parent and which the offspring.

As the Lamarckians say, "If your parents didn't have children, chances are you won't either".

 

[Ygggdrasil]

Clearly both are possible, however, which do you think is more likely?

Assuming only two possible alleles for the gene, the Aa heterozygote has only a 50% chance of contributing allele A, whereas the AA homozygote has a 100% chance of contributing allele A. The other parent can be AA, Aa, or aa, if the heterozygote is the offspring, but if the homozygote is the offspring, neither parent can be aa; the other parent would have to be either AA or Aa $-$ filling out the table, it seems to me, would show more cases in which the AA was the parent.

Let's approach the problem mathematically. Assume a loci with only two alleles, the major allele A and a minor allele a. What is the likelihood of a AA individual having an Aa child p(AA x __ --> Aa). For the three possible genotypes of the unknown parent, this probability is zero if the unknown parent is AA, 1/2 if the unknown parent is Aa, and 1 if the unknown parent is aa. Thus p(obs|AA parent) = (1/2)f(Aa) + f(aa), where f(Aa) is the prevalence of Aa genotypes in the population and f(aa) is the prevalence of aa genotypes in the population. Similarly, we can calculate the probability that an Aa individual has an AA child p(Aa x __ --> AA) = (1/2)f(AA) + (1/4)f(Aa).

If we assume that the population is in Hardy-Weinberg equilibrium, with minor allele fraction q for the a allele, then f(AA) = (1-q)², f(Aa) = 2q(1-q), and f(aa) = q², and we can simplify the expression above:
p(AA x __ --> Aa) = q
p(Aa x __ --> AA) = (1-q)/2

Thus, the likelihood of whether AA is the parent or Aa is the parent depends on the minor allele frequency. If a is more rare, it is more likely that Aa is the parent, but if a is more common, then it is more likely that AA is the parent. Applying Bayes' Law and an uninformative prior, we can get the following graph of the probability that AA is the parent vs the probability that Aa is the parent:

 

[sysprog]

Let's approach the problem mathematically. Assume a loci with only two alleles, the major allele A and a minor allele a. What is the likelihood of a AA individual having an Aa child p(AA x __ --> Aa). For the three possible genotypes of the unknown parent, this probability is zero if the unknown parent is AA, 1/2 if the unknown parent is Aa, and 1 if the unknown parent is aa. Thus p(obs|AA parent) = (1/2)f(Aa) + f(aa), where f(Aa) is the prevalence of Aa genotypes in the population and f(aa) is the prevalence of aa genotypes in the population. Similarly, we can calculate the probability that an Aa individual has an AA child p(Aa x __ --> AA) = (1/2)f(AA) + (1/4)f(Aa).

If we assume that the population is in Hardy-Weinberg equilibrium, with minor allele fraction q for the a allele, then f(AA) = (1-q)², f(Aa) = 2q(1-q), and f(aa) = q², and we can simplify the expression above:
p(AA x __ --> Aa) = q
p(Aa x __ --> AA) = (1-q)/2

Thus, the likelihood of whether AA is the parent or Aa is the parent depends on the minor allele frequency. If a is more rare, it is more likely that Aa is the parent, but if a is more common, then it is more likely that AA is the parent. Applying Bayes' Law and an uninformative prior, we can get the following graph of the probability that AA is the parent vs the probability that Aa is the parent:
View attachment 290020

It seems to me that if A is as likely as a (which had been my working assumption, temporarily disregarding dominant versus recessive traits), your graph shows AA to be more likely to be the parent. You present a to-me-interesting analysis.