AntiRetroScan 1.0 to 1.7
(last updated February 10, 2007)
This documentation refers to AntiRetroScan versions 1.0 to 1.7. These were used from April 2001 to December 2008. The version currently in use is 2.x.
Principle of the method
AntiRetroScan was developed at the University of Siena HIV Monitoring Service (SHMS) in April 2001. Initially, it was just an off-line method to run the same algorithm used at the University of Stanford and store the results in a local database. Starting with the end of the same year, AntiRetroScan has been periodically updated independently from the Stanford algorithm it was derived from.
Similar to the Stanford system, AntiRetroScan analyzes the submitted sequence and extracts the mutations matching those considered to be relevant for drug resistance in the updated reference file. The same file has a score for each of these mutation for each antiretroviral drug, based on the expected role of the mutation itself for that drug. The sum of the scores obtained generates a total score and assigns each drug to one of five susceptibility category defined by different ranges of values.
Unlike the Stanford system, AntiRetroScan also includes a set of remodelling rules changing the total score for a drug in the presence of defined conditions. This approach is aimed at dealing possible synergies among mutation sets which may be not correctly considered with the simple addition of scores.. Thus, AntiRetroScan is a hybrid system, different from the algorithms using only scores (e. g. Stanford) or only rules (e. g. Bayer Trugene, ANRS, Retrogram).
AntiRetroScan output
AntiRetroScan returns five predicted susceptibility levels:
1.     NONE
No resistance mutations found for the drug considered
The drug is fully effective
The sequence contains only accessory mutations or mutations conferring resistance to other drugs but only minimal cross-resistance to the drug considered
The response to the drug is highly likely
3.     PARTIAL
Resistance to the drug considered has started to develop or there are mutations conferring resistance to other drugs with partial cross-resistance to the drug considered
The response to the drug is still possible but further development of resistance is likely
The resistance mutation pattern has a notable impact on susceptibility to the drug considered due to primary mutations selected by the drug itself or extensive cross-resistance caused by other drugs
The response to the drug is low to irrelevant
5.     HIGH
There is complete resistance or cross-resistance to the drug considered
The response to the drug is fully compromised
Sources for definition of scores and rules
The scores defined for each drug-mutation pair and the remodelling rules are derived and periodically updated from the following sources:
Correlation between genotype and in vitro susceptibility. These data are generated by in vitro analysis of laboratory or clinical HIV variants. This kind of information is normally derived from initial studies during the development of the drug and is the necessary starting point for further characterization of the resistance profile for a given drug. However, in vitro susceptibility assays are not devoid of shortcomings, e. g. it may be difficult to measure low resistance levels that are relevant in the clinic for some drugs or it may happen that mutations observed in vitro are extremely uncommon in vivo.
Correlation between genotype and response to treatment in vivo.. This information derives from clinical trials or observational studies and is presently considered the most important source for genotype interpretation algorithms. However, it is often difficult and prone to incorrect conclusions to obtain rules for individual drugs from analysis of data collected from combination therapy studies. Small-sized monotherapy pilot studies and ‘add-on’ studies, where a single drug is added to a background regimen, may provide useful data in this context.
Correlation between genotype and treatment history. This kind of analysis can be performed only on reasonably large databases. The association between treatment and variation in the HIV genome can be evaluated both in cross-sectional and longitudinal studies, provided that multiple sequences from the same patients at different time points are available. This approach may reveal novel associations and improve knowledge on the role of specific mutations. However, as antiretroviral treatment history is often complex, particularly in highly experienced patients, a variable proportion of mutations may have been selected by previous regimens but maintained by other drugs in the current regimen.
Expected results
AntiRetroScan (version 1.4 dated October 12, 2002) has been evaluated in parallel with the on-line system available at the University of Stanford (HIVdb , August 2002 version) and the system included in the Trugene HIV-1 Genotyping Test (Visible Genetics/Bayer, rules 6.0, February 2002). The three methods were used to interpret more than 1,300 sequences with known total of more than 9,000 measured (Antivirogram TM or PhenosenseTM) or inferred (VirtualPhenotypeTM) in vitro phenotypes. AntiRetroScan, Trugene and HIVdb yielded major discordances (susceptible vs. resistant and vice versa) in 8.0%, 10.9% and 12.7% of cases with respect to real phenotype and 2.7%, 5.6% and 6.6% of cases with respect to virtual phenotype.
AntiRetroScan has not yet been evaluated for prediction of virological response to treatment.
For further details you may want to consult the following documents:
Zazzi M, Romano L, Venturi G, Valensin PE. Use of an in-house Windows-based desktop program for prediction of drug susceptibility from HIV genotype. First European HIV Drug Resistance Workshop. Luxembourg, March 6-8, 2003. (pdf poster)
Zazzi M, Romano L, Venturi G, Shafer RW, Reid C, Dal Bello F, Parolin C, Palů G, Valensin PE. Comparative evaluation of three computerized algorithms for prediction of antiretroviral susceptibility from human immunodeficiency virus type 1 genotype. J Antimicrob Chemother 2004;53:356-60. (MEDLINE)
- 02/04/2001. Version 1.0. First use of AntiRetroScan as a tool independent from Stanford HIVdb.
- 01/06/2001. Version 1.1. New rule for the effect of the association between TAMs and M184V on abacavir.
- 23/07/2001. Version 1.2. Updated scores for several mutations for the NRTIs, amprenavir and lopinavir/ritonavir.
- 29/01/2002. Version 1.3. Updated scores for some mutations (69D, 118I, 219R) for some NRTIs (abacavir, didanosine, zalcitabine). Inclusion of tenofovir. Changes in the rules for the lopinavir/ritonavir mutations set.
- 12/10/2002. Version 1.4. Updated scores for several mutations for RTIs and PIs (particularly stavudine, tenofovir, lopinavir/ritonavir). Changes in several rules involving different drugs.
- 02/11/2003. Version 1.5. Updated scores for several mutations for RTIs and PIs (particularly tenofovir, lopinavir/ritonavir). Inclusion of atazanavir. Novel comments on re-sensitizing effects for K65R (zidovudine), L74V/I (zidovudine, tenofovir), Y181C/I (zidovudine, tenofovir), M184V/I (zidovudine, stavudine, tenofovir).
- 05/10/2004. Version 1.5.1. New scoring at RT codon 69 when an insertion is present together with a resistance mutation at codon 69 itself.
- 02/01/2006. Version 1.6. Inclusion of tipranavir. Inclusion of new resistance mutations as derived from genotype-treatment correlation in multiple independent databases and/or included in reference "mutation scores" as indicated by IAS (e. g. ATV or TPV). Acknowledgement of a major role of the TAM #1 pattern in NRTI cross-resistance. Minor changes for all drugs.
- 10/02/2007. Version 1.7. Inclusion of darunavir. Update of the tipranavir algorithm. Minor changes for some mutations.
Prerequisites and limitation of the system
AntiRetroScan accepts an HIV-1 protease and/or reverse transcriptase sequence provided that at least protease codons 10-93 and/or reverse transcriptase codons 41-219 are included. If this relevant portion is incomplete the program warns and stops execution. Protease and reverse transcriptase can be examined separately, however.

The reference sequence used for comparison with the submitted sequence is the subtype B consensus as available at the Los Alamos National Laboratory. The reverse transcriptase region beyond amino acid 400 is not considered.

The system recognizes HIV-1 protease and reverse transcriptase based on sequence scanning and detection of conserved regions. However, at present there is not a true alignment algorithm. Therefore, once the start and end of the sequence has been defined, possible insertions and/or deletions (true or erroneously generated by your sequence analysis) result in reading frameshift(s) and yield misleading interpretation. One- or two-amino acid insertions at reverse transcriptase codon 69, as well as one-amino acid deletions at reverse transcriptase codon 67 or 69 are specifically searched for and thus recognized. These account for virtually all the documented cases of true insertions/deletions in HIV-1 reverse transcriptase. All others insertions/deletions result in frameshift and detection of successive mutations and AntiRetroScan warns of a possible frameshift. All non-alphabetic characters (e. g. * /) in the sequence are removed and this may also generate a frameshift if these characters were intended to mark unread bases. Additional warnings indicative of possible sequence quality problems include:
- the presence of stop codons (TAA, TGA, TAG)
- the presence of three-base degenerate codons (B, D, H, V), however considered as true by the system
- the presence of four-base degenerate codons (N), converted to the wild type base by the system
Subtype analysis is simply made by calculating the percentage of homology between the submitted sequence and the subtype and circulating recombinant form reference sequences as updated at the Los Alamos National Laboratory. Please note that while this approach is rapid and reasonably correct with ‘pure’ subtypes, recombinant forms should be subjected to formal phylogenetic analysis, if possible also including other regions of the same virus genome.
Resistance to antiretroviral drugs is a rapid evolving field, particularly with most recent compounds. The necessary updates of the system thus imply that the same genotype may be interpreted differently in different time periods (i. e. algorithm versions).
Contact us
Address your questions or comments to Maurizio Zazzi.