Consensus-based clock synchronisation (CCS) protocols have gained recent attention in wireless sensor networks. However, the well-known and state-of-the-art protocols are ‘all node based’, that is, every node iterates the consensus algorithm to reach to the synchronised state by exchanging synchronisation messages with the neighbours. This increases the congestion in the network because of extensive message exchanges and induces packet losses and delay in the network. Hence, it is desirable that a subset of connected sensors along with a balanced number of neighbouring sensors should be selected to form a logical topology which will serve as a virtual backbone for the CCS algorithm. This will minimise the overall message complexity and energy consumption in the network as well as balances and minimises delay for faster consensus convergence with optimal synchronisation error. This problem is claimed to be a generalisation of Load Balanced Connected Dominating Set problem which is recently proved to be NP-complete. To make the problem tractable, a genetic algorithm-based strategy is proposed to select the synchronising nodes to form an optimal logical topology.