New models of massive parallel computation devices are suggested. The basic idea of these devices is the modelling of the evolution of a family of cells. It is a modification of DNA-computers or neural networks or abstract chemistry machines. Three versions of the device are defined, the first one computes subsets of natural numbers, the second (or third) type of the device computes subsets of strings -- languages -- using rewriting (or splicing) rules. It is proved that all suggested models compute exactly recursive enumerable sets, even strong restricted conditions on them are given. Some open problems concerning effectiveness of computations are formulated.