Tobias Sing

ROCR: visualizing classifier performance in R

UNLABELLED ROCR is a package for evaluating and visualizing the performance of scoring classifiers in the statistical language R. It features over 25 performance measures that can be freely combined to create two-dimensional performance curves. Standard methods for investigating trade-offs between specific performance measures are available within a uniform framework, including receiver operating characteristic (ROC) graphs, precision/recall plots, lift charts and cost curves. ROCR integrates tightly with Rs powerful graphics capabilities, thus allowing for highly adjustable plots. Being equipped with only three commands and reasonable default values for optional parameters, ROCR combines flexibility with ease of usage. AVAILABILITY http://rocr.bioinf.mpi-sb.mpg.de. ROCR can be used under the terms of the GNU General Public License. Running within R, it is platform-independent. CONTACT tobias.sing@mpi-sb.mpg.de.

Current V3 genotyping algorithms are inadequate for predicting X4 co-receptor usage in clinical isolates

Objective:Integrating CCR5 antagonists into clinical practice would benefit from accurate assays of co-receptor usage (CCR5 versus CXCR4) with fast turnaround and low cost. Design:Published HIV V3-loop based predictors of co-receptor usage were compared with actual phenotypic tropism results in a large cohort of antiretroviral naive individuals to determine accuracy on clinical samples and identify areas for improvement. Methods:Aligned HIV envelope V3 loop sequences (n = 977), derived by bulk sequencing were analyzed by six methods: the 11/25 rule; a neural network (NN), two support vector machines, and two subtype-B position specific scoring matrices (PSSM). Co-receptor phenotype results (Trofile Co-receptor Phenotype Assay; Monogram Biosciences) were stratified by CXCR4 relative light unit (RLU) readout and CD4 cell count. Results:Co-receptor phenotype was available for 920 clinical samples with V3 genotypes having fewer than seven amino acid mixtures (n = 769 R5; n = 151 X4-capable). Sensitivity and specificity for predicting X4 capacity were evaluated for the 11/25 rule (30% sensitivity/93% specificity), NN (44%/88%), PSSM(sinsi) (34%/96%), PSSM(x4r5) (24%/97%), SVMgenomiac (22%/90%) and SVMgeno2pheno (50%/89%). Quantitative increases in sensitivity could be obtained by optimizing the cut-off for methods with continuous output (PSSM methods), and/or integrating clinical data (CD4%). Sensitivity was directly proportional to strength of X4 signal in the phenotype assay (P < 0.05). Conclusions:Current default implementations of co-receptor prediction algorithms are inadequate for predicting HIV X4 co-receptor usage in clinical samples, particularly those X4 phenotypes with low CXCR4 RLU signals. Significant improvements can be made to genotypic predictors, including training on clinical samples, using additional data to improve predictions and optimizing cutoffs and increasing genotype sensitivity.

Determining Human Immunodeficiency Virus Coreceptor Use in a Clinical Setting: Degree of Correlation between Two Phenotypic Assays and a Bioinformatic Model

ABSTRACT Two recombinant phenotypic assays for human immunodeficiency virus (HIV) coreceptor usage and an HIV envelope genotypic predictor were employed on a set of clinically derived HIV type 1 (HIV-1) samples in order to evaluate the concordance between measures. Previously genotyped HIV-1 samples derived from antiretroviral-naïve individuals were tested for coreceptor usage using two independent phenotyping methods. Phenotypes were determined by validated recombinant assays that incorporate either an ∼2,500-bp (“Trofile” assay) or an ∼900-bp (TRT assay) fragment of the HIV envelope gp120. Population-based HIV envelope V3 loop sequences (∼105 bp) were derived by automated sequence analysis. Genotypic coreceptor predictions were performed using a support vector machine model trained on a separate genotype-Trofile phenotype data set. HIV coreceptor usage was obtained from both phenotypic assays for 74 samples, with an overall 85.1% concordance. There was no evidence of a difference in sensitivity between the two phenotypic assays. A bioinformatic algorithm based on a support vector machine using HIV V3 genotype data was able to achieve 86.5% and 79.7% concordance with the Trofile and TRT assays, respectively, approaching the degree of agreement between the two phenotype assays. In most cases, the phenotype assays and the bioinformatic approach gave similar results. However, in cases where there were differences in the tropism results, it was not clear which of the assays was “correct.” X4 (CXCR4-using) minority species in clinically derived samples likely complicate the interpretation of both phenotypic and genotypic assessments of HIV tropism.

Computational methods for the design of effective therapies against drug resistant HIV strains

The development of drug resistance is a major obstacle to successful treatment of HIV infection. The extraordinary replication dynamics of HIV facilitates its escape from selective pressure exerted by the human immune system and by combination drug therapy. We have developed several computational methods whose combined use can support the design of optimal antiretroviral therapies based on viral genomic data.

Involvement of Novel Human Immunodeficiency Virus Type 1 Reverse Transcriptase Mutations in the Regulation of Resistance to Nucleoside Inhibitors

ABSTRACT We characterized 16 additional mutations in human immunodeficiency virus type 1 (HIV-1) reverse transcriptase (RT) whose role in drug resistance is still unknown by analyzing 1,906 plasma-derived HIV-1 subtype B pol sequences from 551 drug-naïve patients and 1,355 nucleoside RT inhibitor (NRTI)-treated patients. Twelve mutations positively associated with NRTI treatment strongly correlated both in pairs and in clusters with known NRTI resistance mutations on divergent evolutionary pathways. In particular, T39A, K43E/Q, K122E, E203K, and H208Y clustered with the nucleoside analogue mutation 1 cluster (NAM1; M41L+L210W+T215Y). Their copresence in this cluster was associated with an increase in thymidine analogue resistance. Moreover, treatment failure in the presence of K43E, K122E, or H208Y was significantly associated with higher viremia and lower CD4 cell count. Differently, D218E clustered with the NAM2 pathway (D67N+K70R+K219Q+T215F), and its presence in this cluster determined an increase in zidovudine resistance. In contrast, three mutations (V35I, I50V, and R83K) negatively associated with NRTI treatment showed negative correlations with NRTI resistance mutations and were associated with increased susceptibility to specific NRTIs. In particular, I50V negatively correlated with the lamivudine-selected mutation M184V and was associated with a decrease in M184V/lamivudine resistance, whereas R83K negatively correlated with both NAM1 and NAM2 clusters and was associated with a decrease in thymidine analogue resistance. Finally, the association pattern of the F214L polymorphism revealed its propensity for the NAM2 pathway and its strong negative association with the NAM1 pathway. Our study provides evidence of novel RT mutational patterns that regulate positively and/or negatively NRTI resistance and strongly suggests that other mutations beyond those currently known to confer resistance should be considered for improved prediction of clinical response to antiretroviral drugs.

Structural Descriptors of gp120 V3 Loop for the Prediction of HIV-1 Coreceptor Usage

HIV-1 cell entry commonly uses, in addition to CD4, one of the chemokine receptors CCR5 or CXCR4 as coreceptor. Knowledge of coreceptor usage is critical for monitoring disease progression as well as for supporting therapy with the novel drug class of coreceptor antagonists. Predictive methods for inferring coreceptor usage based on the third hypervariable (V3) loop region of the viral gene coding for the envelope protein gp120 can provide us with these monitoring facilities while avoiding expensive phenotypic tests. All simple heuristics (such as the 11/25 rule) as well as statistical learning methods proposed to date predict coreceptor usage based on sequence features of the V3 loop exclusively. Here, we show, based on a recently resolved structure of gp120 with an untruncated V3 loop, that using structural information on the V3 loop in combination with sequence features of V3 variants improves prediction of coreceptor usage. In particular, we propose a distance-based descriptor of the spatial arrangement of physicochemical properties that increases discriminative performance. For a fixed specificity of 0.95, a sensitivity of 0.77 was achieved, improving further to 0.80 when combined with a sequence-based representation using amino acid indicators. This compares favorably with the sensitivities of 0.62 for the traditional 11/25 rule and 0.73 for a prediction based on sequence information as input to a support vector machine and constitutes a statistically significant improvement. A detailed analysis and interpretation of structural features important for classification shows the relevance of several specific hydrogen-bond donor sites and aliphatic side chains to coreceptor specificity towards CCR5 or CXCR4. Furthermore, an analysis of side chain orientation of the specificity-determining residues suggests a major role of one side of the V3 loop in the selection of the coreceptor. The proposed method constitutes the first approach to an improved prediction of coreceptor usage based on an original integration of structural bioinformatics methods with statistical learning.

Estimating HIV Evolutionary Pathways and the Genetic Barrier to Drug Resistance

BACKGROUND The evolution of drug-resistant viruses challenges the management of human immunodeficiency virus (HIV) infections. Understanding this evolutionary process is important for the design of effective therapeutic strategies. METHODS We used mutagenetic trees, a family of probabilistic graphical models, to describe the accumulation of resistance-associated mutations in the viral genome. On the basis of these models, we defined the genetic barrier, a quantity that summarizes the difficulty for the virus to escape from the selective pressure of the drug by developing escape mutations. RESULTS From HIV reverse-transcriptase sequences that had been obtained from treated patients, we derived evolutionary models for zidovudine, zidovudine plus lamivudine, and zidovudine plus didanosine. The genetic barriers to resistance to zidovudine, stavudine, lamivudine, and didanosine, for the above 3 regimens, were computed and analyzed. We found both the mode and the rate of development of resistance to be heterogeneous. The genetic barrier to zidovudine resistance was increased if lamivudine was added to zidovudine but was decreased for didanosine. The barrier to lamivudine resistance was maintained with zidovudine plus didanosine, whereas the barrier to didanosine resistance was reduced most with zidovudine plus lamivudine. CONCLUSION Mutagenetic trees provide a quantitative picture of the evolution of drug resistance. The genetic barrier is a useful tool for design of effective treatment strategies.

Characterization and Structural Analysis of Novel Mutations in Human Immunodeficiency Virus Type 1 Reverse Transcriptase Involved in the Regulation of Resistance to Nonnucleoside Inhibitors

ABSTRACT Resistance to antivirals is a complex and dynamic phenomenon that involves more mutations than are currently known. Here, we characterize 10 additional mutations (L74V, K101Q, I135M/T, V179I, H221Y, K223E/Q, and L228H/R) in human immunodeficiency virus type 1 (HIV-1) reverse transcriptase which are involved in the regulation of resistance to nonnucleoside reverse transcriptase inhibitors (NNRTIs). These mutations are strongly associated with NNRTI failure and strongly correlate with the classical NNRTI resistance mutations in a data set of 1,904 HIV-1 B-subtype pol sequences from 758 drug-naïve patients, 592 nucleoside reverse transcriptase inhibitor (NRTI)-treated but NNRTI-naïve patients, and 554 patients treated with both NRTIs and NNRTIs. In particular, L74V and H221Y, positively correlated with Y181C, were associated with an increase in Y181C-mediated resistance to nevirapine, while I135M/T mutations, positively correlated with K103N, were associated with an increase in K103N-mediated resistance to efavirenz. In addition, the presence of the I135T polymorphism in NNRTI-naïve patients significantly correlated with the appearance of K103N in cases of NNRTI failure, suggesting that I135T may represent a crucial determinant of NNRTI resistance evolution. Molecular dynamics simulations show that I135T can contribute to the stabilization of the K103N-induced closure of the NNRTI binding pocket by reducing the distance and increasing the number of hydrogen bonds between 103N and 188Y. H221Y also showed negative correlations with type 2 thymidine analogue mutations (TAM2s); its copresence with the TAM2s was associated with a higher level of zidovudine susceptibility. Our study reinforces the complexity of NNRTI resistance and the significant interplay between NRTI- and NNRTI-selected mutations. Mutations beyond those currently known to confer resistance should be considered for a better prediction of clinical response to reverse transcriptase inhibitors and for the development of more efficient new-generation NNRTIs.

Characterization of novel HIV drug resistance mutations using clustering, multidimensional scaling and SVM-Based feature ranking

We present a case study on the discovery of clinically relevant domain knowledge in the field of HIV drug resistance. Novel mutations in the HIV genome associated with treatment failure were identified by mining a relational clinical database. Hierarchical cluster analysis suggests that two of these mutations form a novel mutational complex, while all others are involved in known resistance-conferring evolutionary pathways. The clustering is shown to be highly stable in a bootstrap procedure. Multidimensional scaling in mutation space indicates that certain mutations can occur within multiple pathways. Feature ranking based on support vector machines and matched genotype-phenotype pairs comprehensively reproduces current domain knowledge. Moreover, it indicates a prominent role of novel mutations in determining phenotypic resistance and in resensitization effects. These effects may be exploited deliberately to reopen lost treatment options. Together, these findings provide valuable insight into the interpretation of genotypic resistance tests.

Molecular Dynamics and Free Energy Studies on the Wild-Type and Mutated HIV-1 Protease Complexed with Four Approved Drugs: Mechanism of Binding and Drug Resistance

The current strategy to improve the quality of life of Human Immunodeficiency Virus (HIV) infected individuals through suppressing viral replication and maintaining the virus at low to undetectable levels is based on highly active antiretroviral therapy (HAART). Protease inhibitors are essential components of most HAART protocols and are often used as the first line of treatment. However, a considerable percentage of new HIV-1 infections are caused by viruses carrying antiretroviral drug-resistant mutations. In this paper molecular dynamics, docking simulations, and free energy analysis of mutated HIV protease complexes were used to estimate the influence of different drug resistance-associated mutations in lopinavir, amprenavir, saquinavir, and atazanavir protease recognition. In agreement with virological and clinical data, the structural analysis showed that the single mutations V82A, I84V, and M46I are associated with higher energetic values for all analyzed complexes with respect to wild-type, indicating their decreased stability. Interestingly, in atazanavir complexes, in the presence of the L76V substitution, the drug revealed a more productive binding affinity, in agreement with hypersusceptibility data.