Rail steels rely primarily on possessing adequate wear and rolling contact fatigue resistance. These properties, together with the toughness, can in principle be optimized by implementing thermomechanical processing assisted by controlled niobium additions. The purpose of the current work is to develop a Nb-microalloying strategy in the context of high-carbon pearlitic and cementite-free bainitic steels. The conventional methods do not leave the critical regions of a rail section in a suitably processed state. An attempt has been made for the first time, to create a pancaked austenite grain structure, with an examination of the consequences on the final product. One of the major difficulties is to ensure that niobium does not segregate during manufacturing, since niobium is a strong carbide former and rail steels traditionally contain large carbon concentrations. Niobium solubility in austenite has been assessed critically and thermodynamic calculations for microsegregation have been taken into account. The aim is to ensure that any primary niobium carbide precipitated from solute-enriched liquid during non-equilibrium solidification, can be taken into solution in austenite during reheating, to mitigate potential effects of coarse precipitates on the final mechanical properties. Rail steels containing 0.01-0.02 wt% Nb have been designed and characterised. In as-cast condition, primary niobium carbides as large as ~10 µm can be observed, which dissolve slowly during reheating. An attempt has been made to develop a model to estimate the dissolution kinetics of the carbides. Dissolved niobium in reheated austenite precipitates during hot deformation as fine niobium carbides (<50 nm) which inhibit austenite recrystallisation by pinning the austenite grain boundaries. Nb-microalloying increases the ‘no-recrystallisation temperature’ of deformed austenite during multi-pass compression tests. The topology of grain deformation has been analysed in terms of stereological calculations and dilatometric experiments have shown that transformation kinetics tend to accelerate when the austenite is deformed below the no-recrystallisation temperature, however the effect is relatively small. The microstructure and mechanical properties of the as-rolled Nb-microalloyed steels have been characterised along with their rolling-sliding wear performance and compared with their non-microalloyed counterparts. Increased austenite grain boundary area and increased dislocation activity due to pancaking, hinder bainite growth which leads to an increased retained austenite volume fraction. This in turn, leads to slightly improved ductility, improved toughness and improved wear resistance in Nb-microalloyed bainitic alloys. Microstructural refinement in Nbmicroalloyed pearlitic alloys does not have any significant effect on tensile and toughness properties, but wear resistance improves significantly. A Bayesian neural network model has been developed to estimate the wear of rails. Predicted trends have been found consistent with metallurgical experience and the perceived noise levels are consistent with reasonable repeatability of the wear testing method used. The model can be applied widely to estimate wear because of its capacity to indicate uncertainty, including both the perceived level of noise in the output, and an uncertainty associated with fitting the function in the local region of input space.