So after learning about the UKDonorLink and their system to NOT contact all potential sibling matches (unless their index is at 99% chance), I started wondering, what really SHOULD be the dividing point between yes and no. The answer really is, it depends on the situation at hand. Since paternity is a black and white answer, most assume that so is siblingship. It's not. It's based on a scale of relatedness from below 1.0 being not related and above 1.0 suggesting relatedness. The closer the score is to 1.0 the more room for ambiguity. The farther away from 1.0 the more likely that the two individuals are either related or not (depending on which direction the score is).
Up until the past few weeks I was unsure what real tried-and-true sibship test results looked like. I knew the higher the index the more likely the two are related, but I assumed there was some sort of maximum (like 100). I had in my mind that scores between certain ranges were either suggestive, inconclusive, or conclusive. It's less-so about the score, but rather the percentage. Scores only slightly over 1.0 receive percentages around 60-70%. A score of 5 is about 83%, a score of 10 is already at 90%. Scores between 10 and 30 are between 90-97% chance of relatedness. Once the scores go past this point they slow down as they reach closer and closer (yet like an asymptote they never reach) 100%. Siblingship indexes can skyrocket way over 100, up into the 1,000s for full siblings - never reaching 100% probability, but nearing closer and closer.
Thus, the probability is what people should be looking at when they see their scores. In the US scores below 90% are considered inconclusive and inadmissible in court. If offspring wish to only accept conclusive results, then anything over 90% would be conclusive that they are in fact siblings. Indexes below this point but above 1.0, the individuals can decide for themselves what they want that result to mean. Some believe that the possibility of two offspring being more distantly related (cousins) as opposed to siblings is unlikely. Yes, it's possible that the donors could have been brothers (offspring would be first cousins). Anything beyond that however (cousins of greater degrees distance) is almost impossible, because even second cousins share only about 4% of their DNA, so the chances of matching on multiple alleles is beyond coincidence.
The importance of having mom's DNA is something I cannot stress enough!! For example, with my new half-sister, according to Damian and I's original analyses (not counting our mom's genotypes) our indexes were between 0.7 (41%) and 1.6 (61%)!! Bring in mom's DNA and our index shot up to 573 (99.8%)!!
So how are siblingship indexes determined???
Siblingship indexes are a measure of the statistical probability of two individuals sharing one or two biological parents. The index is the odds of the individuals being the alleged relation. An index of 2.0 for example, means the odds of the two people being related as 1 in 2. An index of 573 means the odds are 1 in 573. An index of 0.01 means the two individuals 100 times more likely of NOT being related (1/0.01).
For each marker (loci) a single-locus index is determined. It is a complicated mathematical equation that is dependent on a) what each individual's genotype is, b) what each individual's mother's genotype is, and c) what the frequency is of the alleles in question.
Here's some examples to explain this.........
Example #1: If mom 1's genotype is AB for a marker and child 1's genotype is AC, then the child must have inherited the A allele from mom and the C allele from dad. If mom 2's genotype is BD and child 2's genotype is BC, then child 2 must have inherited allele B from mom and allele C from dad. These two children would produce a match based on both alleles HAVING to have been inherited from dad, and the index would be based on the frequency of allele C.
Example #2: If mom 1's genotype is again AB and child 1's genotype is also AB, then it cannot be determined which allele was inherited from mom and which from dad. If mom 2's genotype is AC and child 2's genotype is BC , with allele C being inherited from dad, then these two children would still match because of allele B, but the index would be significantly lower than the index in example #1 because it must take into account the fact that there is only 50% chance of child 1 inheriting allele B from dad. This index is based on the frequency of both allele A and B and the probability of these two children sharing allele B.
Example #3: If mom 1's genotype is AB and so is child 1, while mom 2's genotype is AC and so is child 2's, then the children both match on allele A, but for neither of them can their mother's allele be excluded. This index is based on the frequencies of alleles A, B and C, and is even more drastically lower than examples #1 and #2.
Example #4: If mom 1's genotype is AB and child 1's genotype is AC, then the allele from the dad is C. If mom 2's genotype is BC and child 2's genotype is AC, then the allele from dad had to have been A. Thus, even though child 1 and 2 both have allele C, it is inherited from opposite parents, and thus the children are not a match on this loci.
A non-match index = 0.5, the prior probability.
Then the indexes for each single-locus are multiplied together and that is the combined siblingship index.
I hope that answers some more questions, and as always please send me an email if you have more questions!
Typo fixed: In Example #4 child 2 should have inherited allele A not allele C from the father.
UPDATE: See 5/22/2010 post (What do my siblingship test results mean?) for information about results.