Daniel Hoffmann

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.

Mtreemix: a software package for learning and using mixture models of mutagenetic trees

SUMMARYnMixture models of mutagenetic trees constitute a class of probabilistic models for describing evolutionary processes that are characterized by the accumulation of permanent genetic changes. They have been applied to model the accumulation of chromosomal gains and losses in tumor development and the development of drug resistance-associated mutations in the HIV genome.Mtreemix is a software package for estimating mutagenetic trees mixture models from observed cross-sectional data and for using these models for predictions. We provide programs for model fitting, model selection, simulation, likelihood computation and waiting time estimation.nnnAVAILABILITYnMtreemix, including source code, documentation, sample data files and precompiled Solaris and Linux binaries, is freely available for non-commercial users at http://mtreemix.bioinf.mpi-sb.mpg.de/

Structural modeling of ataxin-3 reveals distant homology to adaptins

Spinocerebellar ataxia type 3 (SCA3) is a polyglutamine disorder caused by a CAG repeat expansion in the coding region of a gene encoding ataxin‐3, a protein of yet unknown function. Based on a comprehensive computational analysis, we propose a structural model and structure‐based functions for ataxin‐3. Our predictive strategy comprises the compilation of multiple sequence and structure alignments of carefully selected proteins related to ataxin‐3. These alignments are consistent with additional information on sequence motifs, secondary structure, and domain architectures. The application of complementary methods revealed the homology of ataxin‐3 to ENTH and VHS domain proteins involved in membrane trafficking and regulatory adaptor functions. We modeled the structure of ataxin‐3 using the adaptin AP180 as a template and assessed the reliability of the model by comparison with known sequence and structural features. We could further infer potential functions of ataxin‐3 in agreement with known experimental data. Our database searches also identified an as yet uncharacterized family of proteins, which we named josephins because of their pronounced homology to the Josephin domain of ataxin‐3. Proteins 2003;50:355–370.

Arevir: a secure platform for designing personalized antiretroviral therapies against HIV

Despite the availability of antiretroviral combination therapies, success in drug treatment of HIV-infected patients is limited. One reason for therapy failure is the development of drug-resistant genetic variants. In principle, the viral genomic sequence provides resistance information and could thus guide the selection of an optimal drug combination. In practice however, the benefit of this procedure is impaired by (1) the difficulty in inferring the clinically relevant information from the genotype of the virus and (2) the restricted availability of this information. We have developed a secure platform for collaborative research aimed at optimizing anti-HIV therapies, called Arevir. A relational database schema was designed and implemented together with a web-based user interface. Our system provides a basis for monitoring patients, decision-support, and computational analyses. Thus, it merges clinical, diagnostic and bioinformatics efforts to exploit genomic and patient therapy data in clinical practice.

Learning multiple evolutionary pathways from cross-sectional data

We introduce a mixture model of trees to describe evolutionary processes that are characterized by the accumulation of permanent genetic changes. The basic building block of the model is a directed weighted tree that generates a probability distribution on the set of all patterns of genetic events. We present an EM-like algorithm for learning a mixture model of K trees and show how to determine K with a maximum likelihood approach. As a case study we consider the accumulation of mutations in the HIV-1 reverse transcriptase that are associated with drug resistance. The fitted model is statistically validated as a density estimator and the stability of the model topology is analyzed. We obtain a generative probabilistic model for the development of drug resistance in HIV that agrees with biological knowledge. Further applications and extensions of the model are discussed.

A New Method for the Fast Solution of Protein-3D-Structures, Combining Experiments and Bioinformatics

Proteins can be considered molecular machines, and protein 3D-structures are key to the understanding of these machines and to many applications in biotechnology and medicine. We are developing a method to speed up the time consuming process of structure determination significantly. The method closely couples bioinformatics for protein structure prediction with fast experiments (chemical cross-linking, specific proteolysis, mass spectrometry) for structure validation. For a given protein, the method iterates over cycles of bioinformatics and experiments to collect more and more information on the protein structure, finally resulting in an Experimentally Validated Model (EVAM) of the structure.

Optimisation of Baseline Genotypic Testing for Safe and Efficient Maraviroc Administration

Different diagnostic parameters may affect the tropism prediction reliability. The impact of usage of FPR cut‐offs<20%, use of viral RNA versus proviral DNA samples, single versus triple amplification, and presence of MVC resistance mutations on tropism prediction at baseline were analysed on 101 patients receiving maraviroc (MVC) and correlated with their clinical outcome. This was a non‐interventional, retrospective study. 82 RNA and 54 DNA samples from the 101 patients receiving MVC were obtained. The V3 region was sequenced and the tropism predicted using the geno2pheno[coreceptor] and T‐CUP tools with FPR cut‐offs of 5%, 7.5%, 10%, 15% and 20%. Additionally, 27/82 RNA and 28/54 DNA samples were analysed in triplicate and 34/82 samples with the ESTA assay. The influence of 16 MVC resistance mutations on clinical outcome was studied. The genotypic susceptibility score (GSS) of the concomitant drugs was mapped to numerical values: susceptible to 1 (or 0.5 for NRTIs), intermediate to 0.5 (0.25 for NRTIs) and resistant to 0. Detection of baseline R5 viruses in RNA (by geno2pheno[coreceptor] and T‐CUP) or DNA (by T‐CUP) samples correlated with MVC‐treatment success. Both tools performed very similarly, with PPVs close to 90%, even with FPR cut‐offs as low as 5%. The use of triple amplification did not improve the prediction value but reduced the number of patients elegible for MVC treatment. No influence of the GSS or MVC resistance mutations on the clinical outcome was detected. Genotypic tropism testing from viral RNA and proviral DNA using the geno2pheno[coreceptor] and T‐CUP systems is valid to select candidates for MVC treatment. Our data suggest that the use of FPR cut‐offs of 5–7.5% and single amplification from RNA or DNA would assure a safe administration of MVC without excluding many patients who could benefit from this potent antiretroviral drug.

A Functional Study on Saposin B and C Using Experimentally Validated Models

To speed up the structure-based analysis of protein function, we have developed the concept of Experimentally Validated 3D Models (EVAMs), i.e., structural models that can be generated relatively fast by a combination of bioinformatics and experiments. Here, we test the suitability of EVAMs for functional studies by applying the concept to two small proteins, saposin B and saposin C. We find that EVAMs are well able to explain some of the functional properties of these proteins, such as their pH-dependent association with lipid bilayers, that are known from independent biophysical experiments.