Is allelic stacking additive?
"Allelic stacking" = the situation when some allele is contributed to
a mixture by two (or more) contributors, hence resulting in a signal
with rfu more than it would be if one of the contributors were removed
(and the remaining contributions were unchanged, i.e. same number of
flourescent molecules).
"Additive" means the that the rfu at such a shared locus would be
the sum of the rfu contributions from the individual contributions.
Otherwise put, additivity means that observed rfu is exactly proportional to quantity.
History and research
It's generally assumed and accepted that stacking is additive
except when very high rfu values overload the detection mechanism.
This additivity is an important assumption because all the
"probabalistic genotyping" programs depend upon it.
It seems intuitively obvious, but in view of the importance of the
assumption it should be checked.
Several years ago I heard that Inman and Rudin had some
experimental data which seemed to contradict additivity, and
that there was some debate between them and STRmix people about
it. I haven't looked at it myself nor learned about any
conclusion. As far as I know it's an open question.
How to test the additivity assumption?
- The simplest experiment, E1,
- would be to
find two people and a locus at which the genotype patterns are PQ for one person
and QR for the other, make a mixture of the two people, and note whether
additivity — rfu(Q) = rfu(P)+rfu(R) —
holds.
But of course that would be a very crude test. It tests
additivity only to the extent that a PQ person would
contribute equal amounts of P as of Q. In practice we are
used to single-source peak height discrepancies of 20% even
for large rfu values.
- Refining the simplistic idea: E2
does E1 many times. Check if additivity holds on average.
- Better but not very good.
- Even on average, the smaller amplicon has a larger
signal. (More efficient PCR duplication.)
- Peak heights are influenced by amplicon sequence.
- Some of an allele's signal is lost to stutter, and propensity to stutter is
affected by sequence.
- Possibly duplication efficiency of the parent amplicon is also
sequence-influenced. I don't think we know.
- E3: Also do many single-source assays of
the PQ and PR genotypes.
- That would give data on idiosyncracies of the particular alleles.
Perhaps with some careful statistics the various awkward complications can be
overcome.