As you all know, there are more than 33 million people living with AIDS worldwide. No cure or vaccine has been unveiled this week in Mexico during the International AIDS Conference. Still, European researchers have developed 'a predictive software system for HIV that could help extend the lives of victims of the killer disease.' The scientists working on the EuResist project have combined HIV databases in Italy, Sweden and Germany, creating what is probably the largest database on AIDS and HIV in the world. Armed with information about more than 18,000 patients, 64,000 therapies, and 240,000 viral mode measurements, the researchers have created new mathematical prediction models, which should soon be available to medical researchers and doctors all over the world. But read more...
The EuResist EU-funded project started in 2006 and officially ended on June 30, 2008. Here is a link to the IST Project Fact Sheet, which says that the total cost of the project was 2.97 million euro. The coordinator was Dr. Francesca Incardona, who works for the Italian company Informa.
Now, let's go back to the ICT Results article mentioned in the introduction of this post to learn more about this project. "By focusing on the genotype of the virus -- information which is inexpensive and easily available -- and combining this with clinical information about the patient, researchers behind the EU-funded EuResist project developed new mathematical prediction models. 'In cases where there is a long history of resistance, this is an indicator of death, so it is important to try all the possibilities,' says Francesca Incardona. For non-crucial cases, this may help reduce the cost of the therapy, by choosing the right combination of drugs that work for the longest time, she suggests."
After combining the HIV databases about patients from Italy, Sweden and Germany, the researchers have developed programs to use all this data. "Medical researchers and doctors can now predict what would happen if a patient with a certain type of virus and certain viral load is given a certain combination of antiretroviral therapy (ARVT). Accurate and reliable prediction of future responses to treatment is based on three basic pieces of information: A quantification of the viral load, a definition of the viral genotype, and the viral load after eight weeks of treatment. The researchers then include other information, such as how the virus is transmitted -- sexual or via a blood transfusion -- the gender of the person, the sexual preference of the person, whether they are drug abusers and other.
Here is a short description of the software tools developed for the EuResist project. "Three different but complementary programmes were developed by the project partners. Each uses the same type of mathematical model to classify a given drug combination as successful or unsuccessful, but is fed with different information. In fact, three approaches are used to extract data from the database to account for different aspects of the disease evolution."
The researchers have decided to make their tools available for free -- at least for the medical community. You can try the Euresist Data Analysis today, but be warned, because the service is still be tested. Academic partners behind this effort include Italy's University of Siena, Sweden's Karolinska Institute, and Germany's Max Planck Institute, all working with Informa.
Besides this, the EuResist team wrote a very interesting technical paper on this research project which has been published in Bioinformatics under the title "Selecting anti-HIV therapies based on a variety of genomic and clinical factors" (Volume 24, Number 13, Pages i399-i406, July 1, 2008). Here are two links to the abstract and to the full paper (PDF format, 8 pages, 209 KB).
In this paper, the members of the team expressed their motivations. "Optimizing HIV therapies is crucial since the virus rapidly develops mutations to evade drug pressure. Recent studies have shown that genotypic information might not be sufficient for the design of therapies and that other clinical and demographical factors may play a role in therapy failure. This study is designed to assess the improvement in prediction achieved when such information is taken into account. We use these factors to generate a prediction engine using a variety of machine learning methods and to determine which clinical conditions are most misleading in terms of predicting the outcome of a therapy."
Here is a short excerpt of the results obtained by the scientists. "Three different machine learning techniques were used: generative–discriminative method, regression with derived evolutionary features, and regression with a mixture of effects. All three methods had similar performances with an area under the receiver operating characteristic curve (AUC) of 0.77. A set of three similar engines limited to genotypic information only achieved an AUC of 0.75. A straightforward combination of the three engines consistently improves the prediction, with significantly better prediction when the full set of features is employed. The combined engine improves on predictions obtained from an online state-of-the-art resistance interpretation system."
For more information, here are links to three papers focused on the "integration of viral genomics with clinical data to predict response to anti-HIV treatment."
- HIV resistance predictor engines specifications (PDF format, 25 pages, 129 KB)
- (PDF format, 29 pages, 782KB)
- Standard Datum (PDF format, 13 pages, 134 KB)
You also should read another article written by another member of the EuResist team, Michal Rosen-Zvi of the IBM Research Laboratory in Haifa, Israel, How we can predict patient response to anti-HIV treatment.
After all these details, let's pause for a minute. These software tools will be available on Internet. Fine. But what about the medications? I think it will be the real test for these tools. Imagine African doctors testing the efficiency of a treatment without being able to prescribe it to their patients -- for financial reasons for example. Please send me your thoughts.
Sources: ICT Results, August 8, 2008; and various websites
You'll find related stories by following the links below.