AFLP-based genome scans are widely used to study the genetics of adaptation and to identify genomic regions potentially under selection. However, this approach usually fails to detect the actual genes or mutations targeted by selection due to the difficulty to obtain the sequences from AFLP fragments. Here we combine classical AFLP outlier detection to 454 sequencing of AFLP fragments obtained from chosen individuals to obtain the sequence of outliers. We applied this approach to the study of resistance to Bacillus thuringiensis israelensis (Bti) toxins in the dengue vector Aedes aegypti. A genome scan of Bti-resistant and Bti-susceptible Aedes aegypti laboratory strains was performed based on 432 AFLP markers. Fourteen outliers were detected using two different population genetics algorithms, a frequentist and a Bayesian approach. Out of these, 11 were successfully sequenced. Three contained transposable elements (TEs) sequences. The 10 outliers that could be mapped at a unique location in the reference genome were located on different supercontigs, underlying the multigenic basis of resistance to Bti toxins. One outlier was in the vicinity of a gene coding for an aminopeptidase potentially involved in Bti toxin binding. Patterns of sequence variability of this gene showed significant deviation from neutrality in the resistant strain but not in the susceptible strain, even after taking into account the known demographic history of the selected strain. This gene is a promising candidate for future functional analysis.