[show abstract][hide abstract] ABSTRACT: Superinfection with drug resistant HIV strains could potentially contribute to compromised therapy in patients initially infected with drug-sensitive virus and receiving antiretroviral therapy. To investigate the importance of this potential route to drug resistance, we developed a bioinformatics pipeline to detect superinfection from routinely collected genotyping data, and assessed whether superinfection contributed to increased drug resistance in a large European cohort of viremic, drug treated patients.
We used sequence data from routine genotypic tests spanning the protease and partial reverse transcriptase regions in the Virolab and EuResist databases that collated data from five European countries. Superinfection was indicated when sequences of a patient failed to cluster together in phylogenetic trees constructed with selected sets of control sequences. A subset of the indicated cases was validated by re-sequencing pol and env regions from the original samples.
4425 patients had at least two sequences in the database, with a total of 13816 distinct sequence entries (of which 86% belonged to subtype B). We identified 107 patients with phylogenetic evidence for superinfection. In 14 of these cases, we analyzed newly amplified sequences from the original samples for validation purposes: only 2 cases were verified as superinfections in the repeated analyses, the other 12 cases turned out to involve sample or sequence misidentification. Resistance to drugs used at the time of strain replacement did not change in these two patients. A third case could not be validated by re-sequencing, but was supported as superinfection by an intermediate sequence with high degenerate base pair count within the time frame of strain switching. Drug resistance increased in this single patient.
Routine genotyping data are informative for the detection of HIV superinfection; however, most cases of non-monophyletic clustering in patient phylogenies arise from sample or sequence mix-up rather than from superinfection, which emphasizes the importance of validation. Non-transient superinfection was rare in our mainly treatment experienced cohort, and we found a single case of possible transmitted drug resistance by this route. We therefore conclude that in our large cohort, superinfection with drug resistant HIV did not compromise the efficiency of antiretroviral treatment.
[show abstract][hide abstract] ABSTRACT: Subtype-dependent selection of HIV-1 reverse transcriptase resistance mutation K65R was previously observed in cell culture and small clinical investigations. We compared K65R prevalence across subtypes A, B, C, F, G and CRF02_AG separately in a cohort of 3076 patients on combination therapy including tenofovir. K65R selection was significantly higher in HIV-1 subtype C. This could not be explained by clinical and demographic factors in multivariate analysis suggesting subtype sequence-specific K65R pathways.
Antimicrobial Agents and Chemotherapy 11/2012; · 4.57 Impact Factor
[show abstract][hide abstract] ABSTRACT: HIV's genetic instability means that sequence similarity can illuminate the underlying transmission network. Previous application of such methods to samples from the United Kingdom has suggested that as many as 86% of UK infections arose outside of the country, a conclusion contrary to usual patterns of disease spread. We investigated transmission networks in the Resina cohort, a 2,747 member sample from Nordrhein-Westfalen, Germany, sequenced at therapy start. Transmission networks were determined by thresholding the pairwise genetic distance in the pol gene at 96.8% identity. At first blush the results concurred with the UK studies. Closer examination revealed four large and growing transmission networks that encompassed all major transmission groups. One of these formed a supercluster containing 71% of the sex with men (MSM) subjects when the network was thresholded at levels roughly equivalent to those used in the UK studies, though methodological differences suggest that this threshold may be too generous in the current data. Examination of the endo- versus exogenesis hypothesis by testing whether infections that were exogenous to Cologne or to Dusseldorf were endogenous to the greater region supported endogenous spread in MSM subjects and exogenous spread in the endemic transmission group. In intravenous drug using group subjects, it depended on viral strain, with subtype B sequences appearing to have origin exogenous to the Resina data, while non-B sequences (primarily subtype A) were almost completely endogenous to their local community. These results suggest that, at least in Germany, the question of endogenous versus exogenous linkages depends on subject group.
Medical Microbiology and Immunology 01/2012; 201(3):259-69. · 3.55 Impact Factor
[show abstract][hide abstract] ABSTRACT: For a long time, the clinical management of antiretroviral drug resistance was based on sequence analysis of the HIV genome followed by estimating drug susceptibility from the mutational pattern that was detected. The large number of anti-HIV drugs and HIV drug resistance mutations has prompted the development of computer-aided genotype interpretation systems, typically comprising rules handcrafted by experts via careful examination of in vitro and in vivo resistance data. More recently, machine learning approaches have been applied to establish data-driven engines able to indicate the most effective treatments for any patient and virus combination. Systems of this kind, currently including the Resistance Response Database Initiative and the EuResist engine, must learn from the large data sets of patient histories and can provide an objective and accurate estimate of the virological response to different antiretroviral regimens. The EuResist engine was developed by a European consortium of HIV and bioinformatics experts and compares favorably with the most commonly used genotype interpretation systems and HIV drug resistance experts. Next-generation treatment response prediction engines may valuably assist the HIV specialist in the challenging task of establishing effective regimens for patients harboring drug-resistant virus strains. The extensive collection and accurate processing of increasingly large patient data sets are eagerly awaited to further train and translate these systems from prototype engines into real-life treatment decision support tools.
[show abstract][hide abstract] ABSTRACT: Etravirine (ETR) is a next generation non-nucleoside reverse transcriptase inhibitor (NNRTI). The studies for ETR EMA approval were almost exclusively performed together with the protease inhibitor (PI) darunavir. However the fact that ETR can be active against NNRTI-pretreated HIV variants and that it is well tolerated suggests its application in PI-free antiretroviral combination therapies. Although approved only for PI-containing therapies, a number of ETR treatments without PIs are performed currently. To evaluate the performance of ETR in PI-free regimens, we analyzed the EURESIST database. We observed a total of 70 therapy switches to a PI-free, ETR containing antiretroviral combination with detectable baseline viral load. 50/70 switches were in male patients and 20/70 in females. The median of previous treatments was 10. The following combinations were detected in the EURESIST database: ETR+MVC+RAL (20.0%); ETR+FTC+TDF (18.6%); 3TC+ETR+RAL (7.1%); 3TC+ABC+ETR (5.7%); other combinations (31.4%). A switch was defined as successful when either ≤50 copies/mL or a decline of the viral load of 2 log10, both at week 24 (range 18-30) were achieved. The overall success rate (SR) was 77% (54/70), and for the different combinations: ETR+MVC+RAL=78.6% (11/14); ETR+FTC+TDF=92.3% (12/13); 3TC+ETR+RAL =80.0% (4/5), 3TC+ABC+ETR=100% (SR 4/4); and for other combinations=67.6% (23/34). These SR values are comparable to those for other therapy combinations in such pretreated patients.
Journal of the International AIDS Society 01/2012; 15(6):18260. · 3.94 Impact Factor
[show abstract][hide abstract] ABSTRACT: Highly active antiretroviral therapy (HAART) has been shown to be effective in many recent trials. However, there is limited data on time trends of HAART efficacy after treatment change.
Data from different European cohorts were compiled within the EuResist Project. The efficacy of HAART defined by suppression of viral replication at 24 weeks after therapy switch was analyzed considering previous treatment modifications from 1999 to 2008. Results: Altogether, 12,323 treatment change episodes in 7,342 patients were included in the analysis. In 1999, HAART after treatment switch was effective in 38.0% of the patients who had previously undergone 1-5 therapies. This figure rose to 85.0% in 2008. In patients with more than 5 previous therapies, efficacy rose from 23.9 to 76.2% in the same time period. In patients with detectable viral load at therapy switch, the efficacy rose from 23.3 to 66.7% with 1-5 previous treatments and from 14.4 to 55.6% with more than 5 previous treatments.
The results of this large cohort show that the outcome of HAART switch has improved considerably over the last years. This result was particularly observed in the context after viral rebound. Thus, changing HAART is no longer associated with a high risk of treatment failure.
[show abstract][hide abstract] ABSTRACT: Sustained suppression of viral replication in HIV-1 infected patients is especially hampered by the emergence of HIV-1 drug resistance. The mechanisms of drug resistance mainly involve mutations directly altering the interaction of viral enzymes and inhibitors. However, protease inhibitors do not only select for mutations in the protease but also for mutations in the precursor Gag and Pol proteins. In this study, we analysed the frameshift-regulating site of HIV-1 subtype B isolates, which also encodes for Gag and Pol proteins, classified as either treatment-naïve (TN) or protease inhibitor resistant (PI-R). HIV-1 Gag cleavage site mutations (G435E, K436N, I437V, L449F/V) especially correlated with protease inhibitor resistance mutations, but also Pol cleavage site mutations (D05G, D05S) could be assigned to specific protease resistance profiles. Additionally, two Gag non-cleavage site mutations (S440F, H441P) were observed more often in HIV-1 isolates carrying protease resistance mutations. However, in dual luciferase assays, the frameshift efficiencies of specific clones did not reveal any effect from these mutations. Nevertheless, two patterns of mutations modestly increased the frameshift rates in vitro, but were not specifically accumulating in PI-resistant HIV-1 isolates. In summary, HIV-1 Gag cleavage site mutations were dominantly selected in PI-resistant HIV-1 isolates but also Pol cleavage site mutations influenced resistance profiles in the protease. Additionally, Gag non-cleavage site mutations accumulated in PI-resistant HIV-1 isolates, but were not related to an increased frameshift efficiency.
Medical Microbiology and Immunology 12/2011; 201(2):213-8. · 3.55 Impact Factor
[show abstract][hide abstract] ABSTRACT: The SnoB study analysed the variability of the integrase (IN) gene of non-B viruses from treatment-naïve patients to determine whether non-B subtypes carry natural resistance mutations to raltegravir (RAL). Plasma viral RNA from 427 patients was gained, and IN sequences were subtyped and screened for subtype-specific highly-variable residues. Seven viruses of different subtypes were phenotypically tested for RAL susceptibility; 359/427 samples could be sequenced. One hundred and seventy samples (47%) were classified as non-B subtypes. No primary RAL resistance-associated mutations (RRAMs) were detected. Certain secondary mutations were found, mostly related to specific non-B subtypes. L74 M was significantly more prevalent in subtype 02_AG, T97A in A and 06_cpx, V151I in 06_cpx, and G163R in 12_BF. Various additional mutations were also detected and could be associated with the subtype too. While K156 N and S230 N were correlated with B subtype, V72I, L74I, T112I, T125A, V201I and T206S were more frequent in certain non-B subtypes. The resistance factors (RF) of 7 viral strains of different subtypes ranged from 1.0 to 1.9. No primary or secondary but subtype-associated additional RRAMs were present. No correlation between RF and additional RRAMs was found. The prevalence of RRAMs was higher in non-B samples. However, the RFs for the analysed non-B subtypes showed lower values to those reported relevant to clinical failure. As the role of baseline secondary and additional mutations on RAL therapy failure is actually not known, baseline IN screening is necessary.
Medical Microbiology and Immunology 04/2011; 200(4):225-32. · 3.55 Impact Factor
[show abstract][hide abstract] ABSTRACT: The EuResist expert system is a novel data-driven online system for computing the probability of 8-week success for any given pair of HIV-1 genotype and combination antiretroviral therapy regimen plus optional patient information. The objective of this study was to compare the EuResist system vs. human experts (EVE) for the ability to predict response to treatment.
The EuResist system was compared with 10 HIV-1 drug resistance experts for the ability to predict 8-week response to 25 treatment cases derived from the EuResist database validation data set. All current and past patient data were made available to simulate clinical practice. The experts were asked to provide a qualitative and quantitative estimate of the probability of treatment success.
There were 15 treatment successes and 10 treatment failures. In the classification task, the number of mislabelled cases was six for EuResist and 6-13 for the human experts [mean±standard deviation (SD) 9.1±1.9]. The accuracy of EuResist was higher than the average for the experts (0.76 vs. 0.64, respectively). The quantitative estimates computed by EuResist were significantly correlated (Pearson r=0.695, P<0.0001) with the mean quantitative estimates provided by the experts. However, the agreement among experts was only moderate (for the classification task, inter-rater κ=0.355; for the quantitative estimation, mean±SD coefficient of variation=55.9±22.4%).
With this limited data set, the EuResist engine performed comparably to or better than human experts. The system warrants further investigation as a treatment-decision support tool in clinical practice.
HIV Medicine 04/2011; 12(4):211-8. · 3.16 Impact Factor
[show abstract][hide abstract] ABSTRACT: The high number of Turkish immigrants in the German state North-Rhine Westphalia (NRW) compelled us to look for HIV-infected patients with Turkish nationality. In the AREVIR database, we found 127 (107 men, 20 women) Turkish HIV patients living in NRW. In order to investigate transmission clusters and their correlation to gender, nationality and self-reported transmission mode, a phylogenetic analysis including pol gene sequences was performed. Subtype distribution and the number of HIV drug resistance mutations in the Turkish patient group were found to be similar to the proportion in the non-Turkish patients. Great differences were observed in self-reported mode of transmission in the heterosexual Turkish male subgroup. Neighbour-joining tree of pol gene sequences gave indication that 59% of these reported heterosexual transmissions cluster with those of men having sex with men in the database. This is the first study analysing HIV type distribution, drug resistance mutations and transmission mode in a Turkish immigrant population.
Medical Microbiology and Immunology 04/2011; 200(4):219-23. · 3.55 Impact Factor
[show abstract][hide abstract] ABSTRACT: Maraviroc (MVC) is the first licensed antiretroviral drug from the class of coreceptor antagonists. It binds to the host coreceptor CCR5, which is used by the majority of HIV strains in order to infect the human immune cells (Fig. 1). Other HIV isolates use a different coreceptor, the CXCR4. Which receptor is used, is determined in the virus by the Env protein (Fig. 2). Depending on the coreceptor used, the viruses are classified as R5 or X4, respectively. MVC binds to the CCR5 receptor inhibiting the entry of R5 viruses into the target cell. During the course of disease, X4 viruses may emerge and outgrow the R5 viruses. Determination of coreceptor usage (also called tropism) is therefore mandatory prior to administration of MVC, as demanded by EMA and FDA. The studies for MVC efficiency MOTIVATE, MERIT and 1029 have been performed with the Trofile assay from Monogram, San Francisco, U.S.A. This is a high quality assay based on sophisticated recombinant tests. The acceptance for this test for daily routine is rather low outside of the U.S.A., since the European physicians rather tend to work with decentralized expert laboratories, which also provide concomitant resistance testing. These laboratories have undergone several quality assurance evaluations, the last one being presented in 2011. For several years now, we have performed tropism determinations based on sequence analysis from the HIV env-V3 gene region (V3). This region carries enough information to perform a reliable prediction. The genotypic determination of coreceptor usage presents advantages such as: shorter turnover time (equivalent to resistance testing), lower costs, possibility to adapt the results to the patients' needs and possibility of analysing clinical samples with very low or even undetectable viral load (VL), particularly since the number of samples analysed with VL < 1000 copies/μl roughly increased in the last years (Fig. 3). The main steps for tropism testing (Fig. 4) demonstrated in this video: Collection of a blood sample Isolation of the HIV RNA from the plasma and/or HIV proviral DNA from blood mononuclear cells Amplification of the env region Amplification of the V3 region Sequence reaction of the V3 amplicon Purification of the sequencing samples Sequencing the purified samples Sequence editing Sequencing data interpretation and tropism prediction.
[show abstract][hide abstract] ABSTRACT: To analyse HIV Gag cleavage site (CS) and non-CS mutations in HIV non-B isolates from patients failing antiretroviral therapy.
Twenty-one HIV isolates were obtained from patients infected with HIV subtype G during an outbreak in Russia 20 years ago. Most patients were failing antiretroviral therapy when genotyping was performed.
HIV Gag CS mutations accumulated in protease inhibitor (PI)-resistant HIV isolates and were correlated with the presence of three or more PI resistance mutations. Only 1 of 11 HIV isolates carrying major protease mutations did not harbour treatment-associated CS mutations. Natural polymorphism 453T, often found in HIV non-B subtypes, seems to favour the selection of CS mutation 453I rather than treatment-associated CS mutation 453L. Resistance-associated non-CS mutations (123E and 200I) were also observed in PI-resistant clinical isolates. Non-CS mutations in the frameshift-regulating site, which controls the synthesis of Gag-Pol, did not affect frameshift efficiency in dual luciferase assays. Of note, one of four HIV isolates from patients failing PI therapies without protease mutations harboured Gag mutations associated with PI resistance (123E and 436R) and reverse transcriptase inhibitor mutations conferring resistance to the backbone drug.
HIV Gag CS mutations commonly occurred in HIV isolates from patients failing PI therapies and natural polymorphisms at the same position influence their emergence. Non-CS mutations previously associated with PI resistance were also observed in clinical isolates. Gag mutations might indicate the evolution of PI resistance even in the absence of protease mutations.
Journal of Antimicrobial Chemotherapy 07/2010; 65(7):1472-6. · 5.34 Impact Factor
[show abstract][hide abstract] ABSTRACT: The success of first-line antiretroviral therapy can be challenged by the acquisition of primary drug resistance. Here we report a case where baseline genotypic resistance testing detected resistance conferring nucleoside/nucleotide reverse transcriptase inhibitor (NRTI)-associated mutations, but no primary mutations for protease inhibitor (PI). Subsequent PI-based HAART with boosted saquinavir led to virological treatment success with persistently undetectable viral load. After treatment simplification from saquinavir to an atazanavir based PI-therapy and no change in backbone therapy rapid virological breakthrough occurred. Retrospective analysis displayed preexisting gag cleavage site mutations which may have reduced the genetic barrier in a clinical relevant manner in combination with the already existing NRTI resistance mutations. Alternatively, this effect could be explained with a different antiviral potency for the respective PIs used.
European journal of medical research 05/2010; 15(5):225-30. · 1.10 Impact Factor
[show abstract][hide abstract] ABSTRACT: Recently, first-line lopinavir failure was observed due to protease mutation 76V. In the present study, we found 76V associated with protease mutation 46I and gag cleavage-site mutation 431V. Longitudinal analysis of patients failing protease inhibitor therapies demonstrated that 76V strictly occurs either together with 46I and/or 431V or in HIV isolates already harbouring one of both mutations. Therefore, all three mutations seem to cooperate in terms of protease inhibitor resistance.
AIDS (London, England) 02/2010; 24(5):779-81. · 4.91 Impact Factor
[show abstract][hide abstract] ABSTRACT: Although genotypic resistance testing (GRT) is recommended to guide combination antiretroviral therapy (cART), funding and/or facilities to perform GRT may not be available in low to middle income countries. Since treatment history (TH) impacts response to subsequent therapy, we investigated a set of statistical learning models to optimise cART in the absence of GRT information.
The EuResist database was used to extract 8-week and 24-week treatment change episodes (TCE) with GRT and additional clinical, demographic and TH information. Random Forest (RF) classification was used to predict 8- and 24-week success, defined as undetectable HIV-1 RNA, comparing nested models including (i) GRT+TH and (ii) TH without GRT, using multiple cross-validation and area under the receiver operating characteristic curve (AUC). Virological success was achieved in 68.2% and 68.0% of TCE at 8- and 24-weeks (n = 2,831 and 2,579), respectively. RF (i) and (ii) showed comparable performances, with an average (st.dev.) AUC 0.77 (0.031) vs. 0.757 (0.035) at 8-weeks, 0.834 (0.027) vs. 0.821 (0.025) at 24-weeks. Sensitivity analyses, carried out on a data subset that included antiretroviral regimens commonly used in low to middle income countries, confirmed our findings. Training on subtype B and validation on non-B isolates resulted in a decline of performance for models (i) and (ii).
Treatment history-based RF prediction models are comparable to GRT-based for classification of virological outcome. These results may be relevant for therapy optimisation in areas where availability of GRT is limited. Further investigations are required in order to account for different demographics, subtypes and different therapy switching strategies.
PLoS ONE 01/2010; 5(10):e13753. · 3.73 Impact Factor
[show abstract][hide abstract] ABSTRACT: We investigated the prevalence of raltegravir resistance-associated mutations at baseline and their evolution during raltegravir therapy in patients infected with different HIV-1 subtypes.
At pre-treatment screening, the integrase gene from plasma samples from patients infected with subtype B and non-B viruses was analysed. Raltegravir resistance evolution was further evaluated in 10 heavily pre-treated patients.
Two hundred and nine plasma samples from 94 subtype B and 115 non-B patients were sequenced. No signature/primary raltegravir resistance mutations were detected at baseline. The secondary mutations L74M, T97A, V151I and G163R were observed with a frequency of <4%. The primary mutations N155H, Q148R/H or Q143R were observed during raltegravir therapy. The Q148R/H was detected only in subtype B. A switch of the primary mutation during raltegravir treatment was not restricted to the subtype B viruses. The prevalence of each primary mutation varied depending on the length of the raltegravir therapy. The Q148R/H was mostly detected after short exposure to raltegravir, while the Y143R was observed only after prolonged raltegravir exposure. We detected an association between the presence of the T206S in the baseline genotype and the absence of the primary Q148R/H mutation or any secondary mutation accompanying the N155H following raltegravir failure.
A number of secondary and additional mutations were found in baseline genotypes. During therapy, when the virus was not optimally suppressed, resistance mutations developed, which were dependent on subtype and time on raltegravir.
Journal of Antimicrobial Chemotherapy 06/2009; 64(1):25-32. · 5.34 Impact Factor
[show abstract][hide abstract] ABSTRACT: Expert-based genotypic interpretation systems are standard methods for guiding treatment selection for patients infected with human immunodeficiency virus type 1. We previously introduced the software pipeline geno2pheno-THEO (g2p-THEO), which on the basis of viral sequence predicts the response to treatment with a combination of antiretroviral compounds by applying methods from statistical learning and the estimated potential of the virus to escape from drug pressure.
We retrospectively validated the statistical model used by g2p-THEO in approximately 7600 independent treatment-sequence pairs extracted from the EuResist integrated database, ranging from 1990 to 2007. Results were compared with the 3 most widely used expert-based interpretation systems: Stanford HIVdb, ANRS, and Rega.
The difference in receiver operating characteristic curves between g2p-THEO and expert-based approaches was significant (P < .001; paired Wilcoxon test). Indeed, at 80% specificity, g2p-THEO found 16.2%-19.8% more successful regimens than did the expert-based approaches. The increased performance of g2p-THEO was confirmed in a 2001-2007 data set from which most obsolete therapies had been removed.
Finding drug combinations that increase the chances of therapeutic success is the main reason for using decision support systems. The present analysis of a large data set derived from clinical practice demonstrates that g2p-THEO solves this task significantly better than state-of-the-art expert-based systems. The tool is available at http://www.geno2pheno.org.
The Journal of Infectious Diseases 03/2009; 199(7):999-1006. · 5.85 Impact Factor
[show abstract][hide abstract] ABSTRACT: The extreme flexibility of the HIV type-1 (HIV-1) genome makes it challenging to build the ideal antiretroviral treatment regimen. Interpretation of HIV-1 genotypic drug resistance is evolving from rule-based systems guided by expert opinion to data-driven engines developed through machine learning methods.
The aim of the study was to investigate linear and non-linear statistical learning models for classifying short-term virological outcome of antiretroviral treatment. To optimize the model, different feature selection methods were considered. Robust extra-sample error estimation and different loss functions were used to assess model performance. The results were compared with widely used rule-based genotypic interpretation systems (Stanford HIVdb, Rega and ANRS).
A set of 3,143 treatment change episodes were extracted from the EuResist database. The dataset included patient demographics, treatment history and viral genotypes. A logistic regression model using high order interaction variables performed better than rule-based genotypic interpretation systems (accuracy 75.63% versus 71.74-73.89%, area under the receiver operating characteristic curve [AUC] 0.76 versus 0.68-0.70) and was equivalent to a random forest model (accuracy 76.16%, AUC 0.77). However, when rule-based genotypic interpretation systems were coupled with additional patient attributes, and the combination was provided as input to the logistic regression model, the performance increased significantly, becoming comparable to the fully data-driven methods.
Patient-derived supplementary features significantly improved the accuracy of the prediction of response to treatment, both with rule-based and data-driven interpretation systems. Fully data-driven models derived from large-scale data sources show promise as antiretroviral treatment decision support tools.
[show abstract][hide abstract] ABSTRACT: Inferring response to antiretroviral therapy from the viral genotype alone is challenging. The utility of an intermediate step of predicting in vitro drug susceptibility is currently controversial. Here, we provide a retrospective comparison of approaches using either genotype or predicted phenotypes alone, or in combination.
Treatment change episodes were extracted from two large databases from the USA (Stanford-California) and Europe (EuResistDB) comprising data from 6,706 and 13,811 patients, respectively. Response to antiretroviral treatment was dichotomized according to two definitions. Using the viral sequence and the treatment regimen as input, three expert algorithms (ANRS, Rega and HIVdb) were used to generate genotype-based encodings and VircoTYPE() 4.0 (Virco BVBA, Mechelen, Belgium) was used to generate a predicted -phenotype-based encoding. Single drug classifications were combined into a treatment score via simple summation and statistical learning using random forests. Classification performance was studied on Stanford-California data using cross-validation and, in addition, on the independent EuResistDB data.
In all experiments, predicted phenotype was among the most sensitive approaches. Combining single drug classifications by statistical learning was significantly superior to unweighted summation (P<2.2x10(-16)). Classification performance could be increased further by combining predicted phenotypes and expert encodings but not by combinations of expert encodings alone. These results were confirmed on an independent test set comprising data solely from EuResistDB.
This study demonstrates consistent performance advantages in utilizing predicted phenotype in most scenarios over methods based on genotype alone in inferring virological response. Moreover, all approaches under study benefit significantly from statistical learning for merging single drug classifications into treatment scores.