NOTE: We asked Dr Doug Gurian-Sherman, a senior scientist at the Union of Concerned Scientists, what he made of the statements by Mark Lynas to the effect that GM is actually safer than conventional breeding because it "just moves a couple of genes, whereas conventional breeding mucks about with the entire genome in a trial and error way." He sent us this response which spells out in detail how Lynas is being both overly simplistic and misleading.
One point that Doug Gurian-Sherman makes that is particularly damaging for the "safer" argument is that GM crops involve not just genetic engineering, which as he notes is less precise in a couple of important ways, but their development into crops that can go into farmers' fields almost invariably requires conventional breeding as well. So genetic engineering isn't an alternative to conventional breeding, but a less precise addition.
Comparing the "precision" of genetic manipulation between genetic engineering and breeding is really like comparing apples and oranges.
GE is clearly less precise in a couple of ways: 1) the gene is inserted into the crop genome randomly (there are technologies in the works that may change this, but have not been used in practice), which can affect the expression of the transgene and surrounding genes, 2) because the new gene may not be co-adapted to the crop genome, it may have unexpected effects on the function of other genes, 3) the tissue culture steps used in the GE process add random genetic and epigenetic changes to the genome (newer methods probably less so, but still some, and these can be reduced by backcrossing).
Typical breeding - crossing of two varieties of a crop, which is the large majority of all breeding, doesn't "mess with" the whole genome. Genes are randomly assorted, which simply means that, as when people have children, copies of co-evolved genes are partitioned in an orderly way among offspring. It is more like shuffling a deck of cards. The genes are mostly the same when varieties of a crop are crossed. The different combinations can have some unpredictable effects, and it is possible that some could be harmful, but that is unlikely.
Importantly, these genes are co-adapted to work together when mixed through breeding, including regulation of gene function. But when GE occurs, with new genes, this probably will not be the case (if it is not "cisgenic") - quite unpredictable results (not necessarily harmful), which one could call imprecise.
That does not mean that conventional breeding is without risk. Harmful recessive genes might be expressed through some crosses during breeding, but if so, they are more likely to make the plant grow poorly than harm anyone. Pro-GE scientists and others often use the examples of breeding using mutagenesis (or tissue culture - one of the steps typically used in GE!), which are unregulated, as examples where there is a lot of imprecision. And that is true. But maybe those things should be regulated rather than deregulating GE! Also, I don't think that either mutagenesis or tissue culture were ever very widely used (there is some dispute about this for mutagenesis. It is hard to get good numbers) because they ARE too messy. On the other hand, about 99 percent of the mutations and so on can typically be removed after mutagenesis or tissue culture by backcrossing to the normal parent for about 10 generations. That is not true, of course, for a transgene!
Sometimes breeding can also activate transposable elements in the genome that move around and cause some mutations. And there are always a few random mutations each generation during breeding. But that is normal, and typically of little or no consequence. In any case, it must be remembered that breeding is always done anyway with GE [breeding is used as a step after the crop is engineered]! After the gene is inserted, that plant is crossed with the parent varieties and other varieties many times. So with GE, we necessarily get the imprecision of GE (insertions site, lack of genomic adaptation), plus the effects of tissue culture, and very minor effects that may occur through breeding. But again, it is important what kind of breeding (or GE, the gene gun is messier than using Agrobacterium) we are talking about.
The other example used by pro-GE people to say that breeding is imprecise is crossing with wild relatives (or even distantly related crop varieties, e.g. landraces), where more genes or alleles (variants of genes, e,g, smooth or wrinkled seed coats in Mendel’s example) can be introduced into the crops. Most of these will be alleles rather than really new types of genes, and all virtually all will be well regulated as part of the crop genome, but that does not mean that some of these will not be harmful. And there are actually a few examples (rolled out ad nauseum by pro-GE scientists, e.g. Lenape or Magnum Bonum potato varieties) of harmful changes. But the genes that can be accessed this way are not nearly as varied as what can be accessed through GE, so I believe that means more uncertainty with GE (as a whole rather than any specific example). For example, the harmful changes in potatoes were increases or changes in alkaloids. But alkaloids already exist in potatoes, and are well known, and so these changes are easy to look for. That is not to say that there may not be some unanticipated harmful changes, just, in my opinion, that they are more likely to be familiar than with GE.
So again, unfortunately, it is not so simple (and, Lynas seems to selectively choose his data to favor GE), and is why I said comparing the precision of GE and breeding is like comparing apples and oranges. For the garden variety of breeding, though, I think it is clear that Lynas is way off, because along with GE itself, breeding is used in the GE process, so there are the additive effects of breeding and GE.
The most challenging examples for those of us supporting breeding are crosses with wild relatives, because there is greater possibility for harm, but also greater opportunity to introduce useful traits.