This paper studies the notion of network slicing in the emerging 5G Time Division Duplex (TDD) networks, considering different applications with diverse service requirements. The proposed solution relies on flexible slicing considering the entire spectrum band, wherein service oriented slices can independently adopt and adjust an UL/DL ratio reflecting traffic conditions. To this end, this paper initially describes and analyses the inter-slice resource allocation problem taking into account different TDD patterns, the slice traffic load and the inter-node interference, which originates by multiple and diverse sources. In this context, a graph-based framework is proposed to reduce the complexity of the problem by decoupling it in two sub-problems: the selection of TDD patterns to find which links should be activated at particular time instances and accordingly perform graph-based resource allocation for all slice traffic collectively. Our performance evaluations show significant enhancements of the UL and DL throughput respectively.