Identifying half-siblings by genetic tests

I have received four related questions from different people. They all concern the issue of half-siblings, and the utility of genetic tests for establishing such a relationship.

1) Is it absolutely possible to tell if two females have the same father? These supposedly half sisters do not share the same mother. The father in question is deceased. Will our DNA reveal, certainly, that we share the same father? How do we find a reliable U.S. company to do this for us?

Information from a company like 23andme will provide you with enough markers to be completely sure. I'm not entirely sure how their interface works, so I can't tell you how easy the interpretation will be, but there will be enough data for you to be sure, and you will be able to get help with the interpretation. The reason is that two women who share a father will have an entire X chromosome in common. That is extremely unlikely to occur otherwise.

2) My sister donated a kidney to me. We were told that we matched 5 out of 6 genetic markers. From that information, is it possible to make an "educated guess" as to whether we are more likely to be full or half sisters? We believed we were half-sisters but some interesting coincidences lead me to believe that we may be full-sisters.

The six markers used for this test are not conclusive. As you know, even half-siblings can be a perfect match. A conclusive test would require many markers. Fortunately, companies like 23andme and Navigenetics provide information about many markers (about 450,000), and those tests could tell you definitively. If you are full siblings, then there will be parts of the genome (about one-fourth of the total) where you are identical for long stretches. That would be extremely unlikely if you are only half siblings. However, a small number of markers (less than 200 or so) would make it harder to make a definitive distinction between being half siblings and being full siblings. Six is definitely too few.

3) I wonder if you could help me. In trying to find my biological father, I came up with what could be two half siblings. The parents in both cases are deceased. I have been given a price of $500 for the three of us to test by saliva. Do you think without any parents, this could prove half siblings or would it be a waste of money?

You will share one of your two alleles with a half-sibling at about half of the sites in your genome. So, the answer is that with enough markers (thousands) the answer will be absolutely clear. The source of DNA (saliva, cheek swab, blood) does not matter much. 23andme will do about 450,000 markers for $400 (per person) and give you lots of additional information. The technology is pretty standard so other firms are probably OK. Just make sure that there are many markers (more than 100,000) and that you get access to the data (not just their interpretation of the data). Once you get your results you'll want to look for large regions of the genome where you and the putative half-sibling share markers. Of course, your putative half-siblings will have to agree to this analysis.

4) I heard that a recent study proved that men don't have half children but any children by the same man are full brothers and sisters irregardless of all different birth mothers. Is there a genetic truth to this?

Two children with the same father and different mothers are referred to as half-siblings. What you are referring to is almost certainly a legal or cultural distinction, not the sort of thing that can be proved by a study.

--------------------------------------
As usual, I invite additional answers in the form of comments. We are looking for answers from people with some expertise, and you will be asked to log in so that we know who you are (no anonymous answers).

A beginner's guide to genetics

Could you please recommend me a few sites on the internet, blogs, or books that's sort of like a beginner's guide to genetics? I would like to learn more about genetics, but I don't exactly know where to begin. One of the books I've read is great, it's called "Genes and DNA" by Charlotte K. Omoto and Paul F. Lurquin. I'm looking for something like that. But most of the other books at the library that I find seem too complicated, and I believe I should begin with the basic stuff first.

Unfortunately, I am not extremely familiar with books of this sort. I provide a link to the Amazon.com reviews of "A Beginner's Guide to Genetics and its Applications." Amazon recommends "Genetics for Dummies" and "Abraham Lincoln's DNA and Other Adventures in Genetics" as similar books. The latter sounds like an interesting read. I also try to recommend informative sites via my Gene Info web site.

I'm posting the question here in the hope that a reader will have suggestions. Although it would be nice to hear from non-experts, I will stick to the policy that while questions can be anonymous, comments (answers) are moderated, and I will only approves comments from people who identify themselves.

Distinguishing full siblings from half siblings

Can you determine if three sisters have the same father by DNA testing of the three sisters?

This is a question about the availability of genetic tests that distinguish full siblings (such as sisters with the same mother and father) from half siblings (such as sisters with the same mother but different fathers). Although full siblings, like parent-child pairs, share half of their alleles, the analysis is a bit trickier. Consider these examples:
The numbers here represents the allelic state of one marker. Each child inherits one allele (one version of the marker) from each parent. In this example, Amy has inherited allele 1 from her father and allele 3 from her mother. Because children must share one allele with each of their parents it is relatively easy to rule out paternity. For example, P2 cannot be Brooke's father because he has completely different alleles from her (1 and 5 vs. 2 and 3). Unlike parents and their children, full siblings can have completely different alleles (e.g. Amy and Casey). Since the question refers to three sisters, it would be possible to rule out all three having the same father (e.g. sisters with genotype 1/3 (Amy), 2/3 (Brooke) and 5/3 (Dawn) cannot all be full sisters because there are three different alleles in combination with the 3, and a single parent can only contribute two possible alleles). However, this is not always the case, and it is even possible for half-sisters to have the same genotype at any one marker (e.g. Amy and Erin) because any two possible fathers could share one or two alleles by chance.

Thus, distinguishing full-siblings from half-siblings comes down to statistics, and "results" are reported as probabilities. Since that is the case, the more markers the better. Many of the genetic testing services probably aren't using enough markers (but they do provide guidance as to how (un)reliable their result is). On the other hand, a full genotyping service, like 23andme, which uses thousands of markers, would provide a reliable result. So, the bottom line is yes, you can determine, with confidence, whether three sisters have the same father by testing the sisters. However, you need to use a lot of markers to do it, and most genetics testing companies probably don't use enough. If you're going to have the testing done, ask the testing service how many markers they use.

What's the best source for information about specific genes?

We are interested in studying gene expressions in living animals, by means of different in-vivo and in-vitro imaging techniques. We also plan to execute tests on living organisms by using Affymetrix or Navigenics or equivalent DNA microarray test chips, but we know that interpretation of such tests are closely related to the ability to execute reliable and effective reannotation vs. existing validated databases of identified genes. We know that services provided by Affymetrix often provide little informative annotation for most known genes (identified as "transcribed locus"), and we would like to know if anyone in the world can provide an effective reannotation service using our DNA microarray test files, i.e. by executing rigorous tests of our Affymetrix chip probeset sequences against all currently available transcript sequences in latest database releases.

This is a question about gene annotation. I think it's really about how to attach a small bit of truly useful information to many genes in a list. The specific question seems to concern Affymetrix data files in particular.

A list of standard sources for information about human genes follows (in most cases I provide a sample link to the page for CFTR). Entries in these databases are extensively linked to each other.

OMIM, Online Mendelian Inheritance in Man. (CFTR)
Entrez Gene, NCBI's searchable database of genes from RefSeq genomes. (CFTR)
GeneCards, a searchable, integrated database of human genes.(CFTR)
HGNC, HUGO gene nomenclature committee, (CFTR)
Ensembl, genome databases for vertebrates and other eukaryotic species. (CFTR)
Ensembl's BioMart tool may be one answer to this question).
GeneWiki, Wikipedia's gene pages. (CFTR; HBB)

How quickly does inbreeding produce homozygosity?

What is the probability that locus is homozygous due to inbreeding after a given number of generations? The answer should be an equation expressing this probability as a function of the number of generations.

This question was inspired by a recent review of mouse genetics (Peters et al. 2007. "The mouse as a model for human biology: a resource guide for complex trait analysis" Nature Reviews Genetics 8, 58-69. doi:10.1038/nrg2025), which states, without qualification, that "Inbred mouse strains are derived from a single parental mating pair with subsequent repeated brother–sister intercrosses and no breeding from non-sibling mice. After 20 generations of inbreeding, the mice are genetically identical and homozygous at all loci." 20 generations is an accepted benchmark that leaves a very low probability of homozygosity, but that probability is not zero. What is it?

This is really three questions (or more).

1) Selfing organisms like Arabidopsis thaliana or Caenorhabditis elegans present a simple case. Here, the probability that a given locus remains heterozygous is simply (1/2)ⁿ, where n is the number of generations.

2) The mouse, which represents all diploid species where crosses between full siblings is possible. Brothers and sisters that share the same two parents. I suspect the equation for this case has been worked out. What is the answer?

3) Species where females store sperm. In this case one can isolate a female each generation. She will have mated with her (possibly half-) brothers prior to isolation. Here the probability of homozygosity is a function not only of the number of generations, but also the probability of the female having mated with a full brother vs. a half-brother.

What do we really know about earlobe genetics?

My husband and I both have attached earlobes. Our newly born son has detached earlobes. How is this possible since attached earlobes are recessive?

- Curious

That is a great question. Many of us learned in high school that attached earlobes are recessive, and it is easy to find that information repeated on the web as an illustration of dominance patterns. An example is the detailed treatment at the Singapore Science Center, which explores various possible parental combinations and concludes that "if both parents were homozygous recessive, they could not have a child with the dominant allele." That is certainly true, and it seems to make a liar out of Curious. Was her child switched in hospital? Should Curious become Alarmed or Suspicious?

I don't think so. I could not find the source of the "fact" that attached earlobes are recessive. I consulted OMIM (the Online Mendelian Inheritance in Man), and they have very little (OMIM 128900). What they do have calls into question the assumption that "attached earlobes are recessive."

Free earlobes are dominant in the view of some. Dutta and Ganguly (1965) suggested polygenic inheritance. There is a variety that is perhaps better classified as 'lobeless' than 'attached.' Lai and Walsh (1966) concluded that 'a simple Mendelian gene effect is unlikely to be responsible for the earlobe types.'

The other issue raised here is whether or not free earlobes is even a bivariate trait (in which case everyone's earlobes could be categorized as either attached or detached). That is necessary for simple dominance to make sense. In the words of students from Willowbrook high school who did a study of this, "it's harder to tell earlobes, so the count on these could be less accurate." OMIM states that "there is a variety that is perhaps better classified as 'lobeless' than 'attached.'" Different forms might show different patterns of inheritance.

I suspect that this, along with a large number of ordinary traits, will be examined anew in the era of personal genomics. For example, the people at deCODE genetics just published a report on the genetics of hair and skin pigmentation (Sulem et al., 2007: Nature Genetics 39:1443, PMID 17952075). Perhaps earlobes will be next.

This question may be most important as a cautionary tale about genetics education. Teaching students about earlobes seems very innocent, but this case illustrates why oversimplification might not be prudent.

What ancestry companies provide actual genetic data?

This question is inspired by an article in today's New York Times business section ("DNA Tests Find Branches but Few Roots"). The article is nice in that it compares the cost of ancestry testing by various companies, shows that results differ, and quotes Henry Louis Gates Jr. making reasoned assessments of the role that DNA testing can play. However, the article fails to say which companies provide customers with the raw genotype information that they can use to make their own assessments as more information becomes publicly available.


This map (from the article) shows the results returned to the author (Ron Nixon) from five different companies. Presumably, the actual genetic information they gathered is a bit less variable than their interpretation. If I paid to have my DNA analyzed, I would want the actual data so that I could make my own interpretations using the best information available from all sources.

The article lists 20 companies that will test your DNA for the purpose of learning about ancestry. Can anyone tell us which ones provide the client with the actual information (sequence or polymorphism data) that can be compared with published genetic demographic data?