In this paper we address the stabilizing leader election problem in the population protocols model augmented with oracles. Population protocols is a recent model of computation that captures the interactions of biological systems. In this model emergent global behavior is observed while anonymous finite-state agents(nodes) perform local peer interactions. Uniform self-stabilizing leader election is impossible in such systems without additional assumptions. Therefore, the classical model has been augmented with the eventual leader detector, Omega?, that eventually detects the presence or absence of a leader. In the augmented model several solutions for leader election in rings and complete networks have been proposed. In this work we extend the study to trees and arbitrary topologies. We propose deterministic and probabilistic solutions. All the proposed algorithms are memory optimal --- they need only one memory bit per agent. Additionally, we prove the necessity of the eventual leader detector even in environments helped by randomization.