A discovery was made which from the outset seemed promising but a little hard to understand. To explain, 'abutting segment boundaries' or ASBs are a type of segment boundary coincidence in which 2 segment matches abut (we first reported 2 such cases here in which the Cr2 8099555 event is accompanied by the additional 8442248 event). Initially we (mistakenly) saw an explanation of ASBs in meiosis, for 4 gametes are produced from one set of parental chromosomes. Complementary gametes show a parent's DNA rejected on one side of a recombination appearing on the other side, thus producing an abutting boundary coincidence which can survive to later generations. With some excitement our presentation 'Fun with Autosomal DNA' used this 'insight' to claim a Baruch Lousada connection across 4 family branches. However Andrew Millard pointed out that the statistics of sperm and egg utilization do not favour our 'insight'. After our first 2 ASBs, new ones were discovered making 20 ASBs in total but then we found 2 more when ELL was added to the set of 12 relatives. For comparison, a set of 13 non-relatives showed increased segment matches (2261 up from 2255), reduced RSBCs (17 not 46), but increased ASBs (46 not 31).

From what follows, it seemed that the terminal SNP of any reported segment normally lies outside the actual segment. For as GEDmatch advised on 24 Oct 2025: 'The boundaries of a segment are practically impossible to get exactly correct. There are alleles in positions which are not SNPs and the crossover is likely not exactly at a SNP. So the mismatch at the front of the segment is before the segment actually starts and the mismatch after the end of the segment is after the segment ends. There is also the possibility that the SNPs after the beginning of the mismatch at the beginning of the segment happen by chance and the segment might actually start after the first aligned SNP following the match. For example if the first SNP inside the segment is AC then it will match what ever is in the other kit but the crossover may be further in. The same is true at the end of the segment. Finally if the two kits are not from the same vendor chip set there are SNPs which do not align and had those SNPs been available the mismatch SNPs which bound the segment may be different'. 

From those comments, it appears that ASBs should be reported as having a 1 or 2 SNP overlap. But they aren't, so - in Qmatch - we infer that GEDmatch reports a segment boundary halfway between the last misaligned and the first aligned SNP. That is, as we belatedly realised, the GEDmatch comments above refer to normal matching, and not Qmatching. After all, Qmatch is a proprietary product the details of which are kept secret and it is behind GEDmatch's (Tier 1) paywall. Qmatch was recommended to us by GEDmatch for use with small matches, and we have indeed found that without it ASBs do not appear (at least readily). In Qmatch it was easy for us to notice ASBs unaided, because the end of one segment is reported as numerically identical to the beginning of the other abutting segment! AI will be needed for sets larger than our 13 relatives to avoid missing any ASBs, and this would also allow a search for ASBs showing small overlaps or misses eg of a few hundred nucleotide positions (should Qmatch methodology change and inadvertently remove exact ASB matching which is currently so beneficial). In fact, even with our 13 person sets, AI showed that ASBs were missed (4 out of 46 for the comparison set of 13, and 9 out of 31 for the 13 relatives).

The following diagram shows in essence what causes ASBs - starting with 3-person ASBs (3pASBs). Thus, where an ancestral crossover in one sibling is carried forward into a present-day descendant and is accompanied (in 2 further present-day descendants) by a stretch of each parent's DNA which bridges the crossover, ASBs can result under favourable circumstances. By this is meant that they will obviously not be detectable in cases where the ancestral parents match in the region surrounding the crossover. In any event, it is important to consider whether the ancestors defining an ASB are 'parents' as shown in the diagram and not immediate ancestors thereof - which may not be a concern for us because our 3cM match threshhold size makes it likely that the ancestral family is no earlier than that of Amador de Lousada (11 generations back):


In any case, our insight allows us to also understand the small number of 4-person ASBs as well - for here, a pair of relatives from the same family branch can act as a surrogate for Relative 1 in the above chart. That is, both of the relatives in the pair carry the ancestral crossover - a situation which of course is less common given the many reproductive events which might have eliminated the crossover of interest.

Of the 22 3pASBs we found, the following chart shows how they fit the model, and how they posed a challenge for us. The challenge was to recognise that 14 ASBs relate not to the ancestral Amador/Briatis ancestral family but to subsequent intra-branch crossovers. The same applies to 4 of the 4pASBs. The real significance of the chart is that our ASBs provide abundant evidence of interbranch connections. Further, we are now able to associate an ancestral sibling with each relative, though we are unable to distinguish Isaac from Abraham. To distinguish Isaac from Abraham will need a DNA sample from the Surinam Baruch Lousadas - who, so our genealogy tells us, descend from Isaac and not Abraham!





We can now see the potential of ASBs, for in the following chart of proven matches, we have added the 21 extra 3pASB connections from the previous chart to our small match procedure. We have also added the 10 extra 4pASB connections. It can be seen that in total these ASB contributions greatly outnumber the 5 matches from RSBCs and the 10 from Qmatch. This large ASB contribution depends on genealogy, for the following reason. Though our comparison set of people generates 50% more ASBs than our set of relatives, this modest figure disguises two dramatic insights. The first is that while our sample of relatives essentially entails just 6 ancestral families - Amador/Briatis, Lichtenstadt, Montefiore, Nunes, Martin and one other, the comparison sample entails 10000 or so ancestral families (given the way ancestor numbers double each generation - remembering we are talking 11 generations). Thus the crossovers for the relatives (speaking unsubtly) are 1000 times as important as those for the comparison set. It is however worth considering whether ASBs may be able, in the absence of genealogical information, to re-construct some lineages within a random sample - if that was of interest to researchers connected to the people in the sample.