In this paper we define and investigate a binary word operation that formalizes an experimentally observed outcome of DNA computations, performed to generate a small gene library, and implemented using a DNA recombination technique called Cross-pairing Polymerase Chain Reaction (XPCR). The word ble nding between two words αwγ1 and γ2wβ that share a non-empty overlap w, results in αwβ. Interestingly, this phenomenon has been observed independently in linguistics, under the name “blend word” or “portmanteau”, and is responsible for the creation of words in the English language such as smog (smoke + fog), labradoodle (labrador + poodle), and Brangelina (Brad + Angelina). Technically, word blending is related to the binary word operation Latin product, the crossover operation, and simple splicing. We study closure properties of the families in the Chomsky hierarchy under word blending, language equations involving this operation, and its descriptional state complexity when applied to regular languages. We also define iterated word blending and show that, for a given alphabet, there are finitely many languages that can be obtained from an initial language by iterated word blending.

We propose and investigate a formal language operation inspired by the naturally occurring phenomenon of DNA primer extension by a DNA-template-directed DNA Polymerase enzyme. Given two DNA strings u and v, where the shorter string v (called primer) is Watson-Crick complementary and can thus bind to a substring of the longer string u (called template) the result of the primer extension is a DNA string that is complementary to a suffix of the template which starts at the binding position of the primer. The operation of DNA primer extension can be abstracted as a binary operation on two formal languages: a template language L 1 and a primer language L 2 . We call this language operation L 1 -directed extension of L 2 and study the closure properties of various language classes, including the classes in the Chomsky hierarchy, under directed extension. Furthermore, we answer the question under what conditions can a given language of target strings be generated from a given template language when the primer language is unknown. We use the canonic inverse of directed extension in order to obtain the optimal solution (the minimal primer language) to this question.