Take a step back and ask if there's any good analog
synthesis, at all.
Digital, you have about 3-4 care-abouts - area, power,
timing closure, functionality.
Analog, you typically have dozens of parametrics per
block as well as noise, matching, stability, input / output
ranges and impedances. Making specification alone, a
big hassle and then you have the plurality of candidate
topologies for sub-blocks and their compatibility issues.
Reuse of the known-good is thus much better in terms
of success probability, than synthesis (for which, you
would need someone to have gone through a whole lot
of infrastructure work - think digital cell library development,
cubed).
Professors and tool developers sure love to talk about it,
though. Maybe you'll find a text. But as they say, don't
believe everything you read.