The probability of point mutations is expected to be highly influenced by the flanking nucleotides that surround them, known as the sequence context. This phenomenon may be mainly attributed to the enzyme that modifies or mutates the genetic material, because most enzymes tend to have specific sequence contexts that dictate their activity. Here, we develop a statistical model that allows for the detection and evaluation of the effects of different sequence contexts on mutation rates from deep population sequencing data. This task is computationally challenging, as the complexity of the model increases exponentially as the context size increases. We established our novel Bayesian method based on sparse model selection methods, with the leading assumption that the number of actual sequence contexts that directly influence mutation rates is minuscule compared with the number of possible sequence contexts. We show that our method is highly accurate on simulated data using pentanucleotide contexts, even when accounting for noisy data. We next analyze empirical population sequencing data from polioviruses and HIV-1 and detect a significant enrichment in sequence contexts associated with deamination by the cellular deaminases ADAR 1/2 and APOBEC3G, respectively. In the current era, where next-generation sequencing data are highly abundant, our approach can be used on any population sequencing data to reveal context-dependent base alterations and may assist in the discovery of novel mutable sites or editing sites.