Abstract: Drug resistance is a major problem in the treatment of Acquired Immune Deficiency Syndrome (AIDS), due to the very high replication and mutation rate of the human immunodeficiency virus (HIV). Identification of mutations associated with drug resistance and understanding their evolutionary dynamics is critical for both individualized treatment selection and new drug design. We have performed automated mutation analysis of HIV-1 protease and reverse transcriptase (RT) from 50,000 HIV-1 positive patient plasma samples sequenced by Specialty Laboratories Inc. from 1999 to 2003. This analysis generated ∼2.3 million mutation data and enabled calculation of statistically significant K_a/K_s values for individual amino acid mutation in protease and RT. Positive selection analysis detected most of the known drug resistance mutations. We also discovered 163 new amino acid mutations in HIV protease and RT that are strong candidates for drug resistance or fitness. Since amino acids at different sites can interact, a mutation at one site can actually change the selection pressure at other sites. To address this problem, we have extended the K_a/K_s concept to consider interactions with other sites, by formulating a conditional K _a/K_s that measures the effects of mutations at one site on the selection pressure for mutations at another site. 412 codon pairs showed strong positive conditional relationships according to the conditional K _a/K_s analysis. Based on the results of this analysis, we generated a rate diagram to show the speed of all the possible evolutionary paths the virus might follow under the combined selection of existing drug treatments. This diagram revealed specific accessory mutations that greatly accelerate the evolution of multi-drug resistance, and other 'kinetic trap' mutations that block it. To assess the robustness of our approach, we have applied it to four completely independent datasets. The K_a/K_s and conditional K_a/K_s results were highly reproducible among the four independent datasets, both qualitatively and quantitatively, suggesting that they are detecting real drug-resistance and viral fitness mutations in the wild HIV-1 population. Conditional K_a/K_s analysis can distinguish primary vs. accessory mutations for many well-studied examples, and K_a/K_s analysis of treated vs. untreated patient data can distinguish drug-resistance vs. viral fitness mutations.