To the left is a panel from a new paper in PLoS Genetics, Selection-Driven Gene Loss in Bacteria. The y-axis is selection, so above 0 represents a positive selection coefficient, and below a negative one. The lineages above the x-axis then are more fit against the baseline wild type (selection coefficients above 0.01 can be considered rather strong, with 0.08 very strong). The top-line result of this paper is that 11 out of 55 deletion constructs in bacterial lineages seem to result in increased fitness. This is rather weird if you are viewing it from a homocentric perspective, or more ecumenically, a multicelluar organismic perspective. If you add copies of a gene the result is not always good because of dosage effects (Down Syndrome is an extreme case of this on the smallest chromosome). If you delete whole regions of the genome the results are usually disastrous. This comports with common sense. If you break something it isn’t usually a good thing. It was there for a reason.
But this heuristic may not be applicable to bacteria. Or, more precisely, when you remove the large set of parameters which constrain the adaptive landscape of multicellular organisms, you may be exploring an evolutionary space with somewhat different rules. If someone reported that 25% of deletions in a multicellular organism resulted in increased fitness I’d probably be curious if there was a basic human error somewhere along the way generating this crazy proportion. But fast-breeding and genomically economically designed bacteria may not be governed by the same expectations.