Application of a new informatics tool for contamination screening in the HIV sequencing laboratory

Background: Current HIV-1 sequencing-based methods for detecting drug resistance-associated mutations are open and susceptible to contamination. Informatic identification of clinical sequences that are nearly identical to one another may indicate specimen-to-specimen contamination or another laboratory-associated issue. Objectives: To design an informatic tool to rapidly identify potential contamination in the clinical laboratory using sequence analysis and to establish reference ranges for sequence variation in the HIV-1 protease and reverse transcriptase regions among a U.S. patient population. Study design: We developed an open-source tool named HIV Contamination Detection (HIVCD). HIVCD was utilized to make pairwise comparisons of nearly 8000 partial HIV-1 pol gene sequences from patients across the United States and to calculate percent identities (PIDs) for each pair. ROC analysis and standard deviations of PID data were used to determine reference ranges for between-patient and within-patient comparisons and to guide selection of a threshold for identifying abnormally high PID between two unrelated sequences. Results: The PID reference range for between-patient comparisons ranged from 83.8 to 95.7% while within-patient comparisons ranged from 96 to 100%. Interestingly, 48% of between-patient sequence pairs with a PID > 96.5 were geographically related. The selected threshold for abnormally high PIDs was 96 (AUC = 0.993, sensitivity = 0.980, specificity = 0.999). During routine use, HIVCD identified a specimen mix-up and the source of contamination of a negative control. Conclusions: In our experience, HIVCD is easily incorporated into laboratory workflow, useful for identifying potential laboratory errors, and contributes to quality testing. This type of analysis should be incorporated into routine laboratory practice.