Computation underlies the genetic regulatory network activities. Previous studies have designed and engineered systems that can perform single logic gate functionalities, trying to avoid external and internal random fluctuations. In this work, we demonstrate the possibility to exploit noise when it cannot be eliminated. In particular, we adapt the LSR paradigm to a single-gene network derived from the bacteriophage λ and to a more robust two-gene network derived from the yeast S. cerevisiae. Our results demonstrate that in both cases there is an optimal amount of noise where the biological logic gate can be externally reprogrammed (i.e. switch from the AND to the OR gate) and perform well according to the truth table.