I’ve just seen mention of a citizen science compatible DNA barcoding kit, for under £2,000, plus the cost of consumables. On the other hand, if I’ve read the manufactor’s site correctly, it does the extraction and amplification, but not the sequencing, so it strikes me as being of less use than Oxford Nanopore’s MinION.
An additional issue is that the primer for plant barcoding is for the rbcL gene, which from my understanding is too conserved for fine scale taxonomy - it’ll do fine for separating genera, but may not be reliable for species. (One of my back burner projects is to collect several hundred rbcL sequences from GenBank and run them through MEGA-X, which would give me a better feel for how using rbcL is for barcoding.)
I am not into this except (Dodd, M. E., Silvertown, J., & Chase, M. W. 1999. Phylogenetic analysis of trait evolution and species diversity variation among angiosperm families. Evolution 53: 732-744 which used rbcl) I thought multiple genes are now used if possible in many groups. Actually you may use different markers for small groups to detect the fine scale differences between closely related species or sub species or individuals.
I guess you might be more interested in these fine scale differences so rbcl may be of no use.
There are huge numbers of papers now on which markers to use for which groups of species. This might call into question somewhat the eDNA systems or rather need to carefully understand what they can and can’t tell about which species are present in the environment. As mentioned elsewhere on the forum there are citizen science eDNA kits too where you do the sampling and the lab does the chemistry.
Parts of the nrDNA array are widely used - especially ITS1, but also ITS2, ETS, and 5.8S RNA. ITS1 didn’t resolve Sidalcea. (Species were not monophyletic on the inferred tree.) nrDNA evolution has some perculiarities. For chloroplasts the early work used rbcL, ndfH and atpB. Nowadays matK and the trnL-trnF IGS are popular, often combined with rbcL. Nowadays it would seem to be possible to throw data at the problem, and use whole chloroplast genomes - but in naive hands those seem to be vulnerable to artefacts, with papers being published with Gossypium and Hibiscus not resolved from each other. (I think the cause is that the orientation of the IR region is evolutionarily labile, and for taxon of the right divergence the essentially random orientation of the region in a sequence outweighs the genuine phylogenetic signal.)
A bit more browing on the site found that they will sequence your prepared DNA samples for you for £7 each. The question which arises is how does the cost of doing the DNA separation with this kit and sending the DNA off, compare with sending off a leaf and letting the service do the DNA preparation. If I don’t have the wrong end of the stick you don’t need the gel electrophoresis stage for barcoding, and I’m not sure that you need the PCR stage either.
Pricewise, I’d compare it with a posh film camera. The cost of consumables likely adds up, while digital cameras have negligible consumable costs - just recharging batteries, and maybe paying for extra storage space.
It would become cost effective over something like 300 to 500 samples or something like that. I base my estimate on the consumables being a little under half the £7 lab fee. Then there is the outbound postage to add.
Looking at the extended warranty fees adds a operational life estimate too.
Compact digital cameras do about 10,000 to 20,000 photos before the lens motors begin to get a bit sloppy. Divide the price by 10,000 and you get the cost per photo.
You will get very tempted to sample everything once you own such a desktop lab.
Now totally out of my depth, but if you are looking within species then there may be simpler (potentially cheaper) techniques than full DNA analysis, allozymes for example. However there may not be automated ways of doing that kind of analysis.
While I’m not strictly qualified to judge I don’t trust RAPD for phylogeny (I suspect that fragment length is rather like chromosome number, with too high a rate of homoplasy for reliable phylogenetic results) I expect it would be good enough to assign cultivars to species and identify hybrids.