Quantcast


CURATOR
A pinboard by
Endre Szvetnik

I cover science and tech news for Sparrho and work with Sparrho Heroes to curate, translate and disseminate scientific research to the wider public.

PINBOARD SUMMARY

ART supresses HIV in infected patients, but due to its mutation, drug resistance is on the rise.

Scientists have improved machine learning algorithms to help finding the right antiretroviral medications for patients who are developing drug resistance.

In 10 seconds? HIV is a highly mutating virus capable of rendering useless the drugs taken by patients. Computer modelling using machine learning is getting better at predicting who is at risk, helping to save lives and money.

What’s the breakthrough? Scientists have developed a computer model to simulate how HIV mutates in people to suggest the best possible treatments. Working with a number of variables, ranging from peoples’ behaviour to how drugs work on HIV mutations, the algorithm predicts which alternative drugs would save more lives and money in the long run.

Why do we need computing power to keep HIV at bay? Because the virus has been adapting against antiretroviral drugs in the past decades. Predicting where in the world this resistance occurs needs a lot of data crunching. A recent WHO study found that in Uganda, almost 16% of newly infected people were resistant to a key class of first-line drugs, usually the doctors’ first choice.

And what do these algorithms do? For example, they predicted that HIV-related deaths could be reduced by administering dolutegravir as a first-line antiretroviral drug in sub-Saharan Africa. The choice is very important as it can affect the future behaviour and mutations of HIV in patients.

How do they do it? They sift through through massive volumes of data about HIV mutations to find the ones that will be drug-resistant. Currently, laboratory testing methods are still widely used for this task. By using algorithms, scientists have managed to improve the prediction rate of the current gold standard resistance prediction test – known as ANRS – by up to 28%.

Are there other applications of algorithms against HIV? Yes, for example, algorithms can quickly determine ‘viral load’ patterns in large groups of people (i.e. how much HIV there is in their bloodstreams). This classifies patients into groups to help assign the right therapies to them.

So can machine learning help to cure HIV? We hope so! We mentioned broadly neutralising antibodies (bNAbs) in an earlier Digest – scientists are using machine learning techniques to build and combine more effective bNAbs. The algorithms look through millions of parameters to seek common patterns or weak points in HIV that can be targeted by a vaccine.


Why AI is good news for medicine

From diagnosing cancers and predicting bipolar episodes to establishing personalised doses, machine learning can do the heavy lifting that is needed for better medical outcomes.

Being able to find patterns in masses of data, it can be used to direct drug development by suggesting ideal candidates for medical trials and help scientists better understand the progression of various diseases, such as MS or diabetes.

Also, machine learning offers an unexpected benefit: it’s good at recognising letters, even doctors’ notoriously bad handwriting.

11 ITEMS PINNED

Predicting response to antiretroviral treatment by machine learning: the EuResist project.

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.

Pub.: 31 Jan '12, Pinned: 10 May '18

HIV-1 fitness cost associated with escape from the VRC01 class of CD4 binding site neutralizing antibodies.

Abstract: Broadly neutralizing antibodies (bNAbs) have been isolated from selected HIV-1-infected individuals and shown to bind to conserved sites on the envelope glycoprotein (Env). However, circulating plasma virus in these donors is usually resistant to autologous isolated bNAbs, indicating that during chronic infection, HIV-1 can escape from even broadly cross-reactive antibodies. Here, we evaluate if such viral escape is associated with an impairment of viral replication. Antibodies of the VRC01 class target the functionally conserved CD4 binding site and share a structural mode of gp120 recognition that includes mimicry of the CD4 receptor. We examined naturally occurring VRC01-sensitive and -resistant viral strains, as well as their mutated sensitive or resistant variants, and tested point mutations in the backbone of the VRC01-sensitive isolate YU2. In several cases, VRC01 resistance was associated with a reduced efficiency of CD4-mediated viral entry and diminished viral replication. Several mutations, alone or in combination, in the loop D or β23-V5 region of Env conferred a high level of resistance to VRC01 class antibodies, suggesting a preferred escape pathway. We further mapped the VRC01-induced escape pathway in vivo using Envs from donor 45, from whom antibody VRC01 was isolated. Initial escape mutations, including the addition of a key glycan, occurred in loop D and were associated with impaired viral replication; however, compensatory mutations restored full replicative fitness. These data demonstrate that escape from VRC01 class antibodies can diminish viral replicative fitness, but compensatory changes may explain the limited impact of neutralizing antibodies during the course of natural HIV-1 infection.Some antibodies that arise during natural HIV-1 infection bind to conserved regions on the virus envelope glycoprotein and potently neutralize the majority of diverse HIV-1 strains. The VRC01 class of antibodies blocks the conserved CD4 receptor binding site interaction that is necessary for viral entry, raising the possibility that viral escape from antibody neutralization might exert detrimental effects on viral function. Here, we show that escape from VRC01 class antibodies can be associated with impaired viral entry and replication; however, during the course of natural infection, compensatory mutations restore the ability of the virus to replicate normally.

Pub.: 30 Jan '15, Pinned: 10 May '18

Fitness landscape of the human immunodeficiency virus envelope protein that is targeted by antibodies

Abstract: HIV is a highly mutable virus, and over 30 years after its discovery, a vaccine or cure is still not available. The isolation of broadly neutralizing antibodies (bnAbs) from HIV-infected patients has led to renewed hope for a prophylactic vaccine capable of combating the scourge of HIV. A major challenge is the design of immunogens and vaccination protocols that can elicit bnAbs that target regions of the virus’s spike proteins where the likelihood of mutational escape is low due to the high fitness cost of mutations. Related challenges include the choice of combinations of bnAbs for therapy. An accurate representation of viral fitness as a function of its protein sequences (a fitness landscape), with explicit accounting of the effects of coupling between mutations, could help address these challenges. We describe a computational approach that has allowed us to infer a fitness landscape for gp160, the HIV polyprotein that comprises the viral spike that is targeted by antibodies. We validate the inferred landscape through comparisons with experimental fitness measurements, and various other metrics. We show that an effective antibody that prevents immune escape must selectively bind to high escape cost residues that are surrounded by those where mutations incur a low fitness cost, motivating future applications of our landscape for immunogen design.

Pub.: 08 Jan '18, Pinned: 10 May '18

Cost-effectiveness of public-health policy options in the presence of pretreatment NNRTI drug resistance in sub-Saharan Africa: a modelling study.

Abstract: There is concern over increasing prevalence of non-nucleoside reverse-transcriptase inhibitor (NNRTI) resistance in people initiating antiretroviral therapy (ART) in low-income and middle-income countries. We assessed the effectiveness and cost-effectiveness of alternative public health responses in countries in sub-Saharan Africa where the prevalence of pretreatment drug resistance to NNRTIs is high.The HIV Synthesis Model is an individual-based simulation model of sexual HIV transmission, progression, and the effect of ART in adults, which is based on extensive published data sources and considers specific drugs and resistance mutations. We used this model to generate multiple setting scenarios mimicking those in sub-Saharan Africa and considered the prevalence of pretreatment NNRTI drug resistance in 2017. We then compared effectiveness and cost-effectiveness of alternative policy options. We took a 20 year time horizon, used a cost effectiveness threshold of US$500 per DALY averted, and discounted DALYs and costs at 3% per year.A transition to use of a dolutegravir as a first-line regimen in all new ART initiators is the option predicted to produce the most health benefits, resulting in a reduction of about 1 death per year per 100 people on ART over the next 20 years in a situation in which more than 10% of ART initiators have NNRTI resistance. The negative effect on population health of postponing the transition to dolutegravir increases substantially with higher prevalence of HIV drug resistance to NNRTI in ART initiators. Because of the reduced risk of resistance acquisition with dolutegravir-based regimens and reduced use of expensive second-line boosted protease inhibitor regimens, this policy option is also predicted to lead to a reduction of overall programme cost.A future transition from first-line regimens containing efavirenz to regimens containing dolutegravir formulations in adult ART initiators is predicted to be effective and cost-effective in low-income settings in sub-Saharan Africa at any prevalence of pre-ART NNRTI resistance. The urgency of the transition will depend largely on the country-specific prevalence of NNRTI resistance.Bill & Melinda Gates Foundation, World Health Organization.

Pub.: 28 Nov '17, Pinned: 10 May '18

Revealing patterns in HIV viral load data and classifying patients via a novel machine learning cluster summarization method

Abstract: HIV RNA viral load (VL) is an important outcome variable in studies of HIV infected persons. There exists only a handful of methods which classify patients by viral load patterns. Most methods place limits on the use of viral load measurements, are often specific to a particular study design, and do not account for complex, temporal variation. To address this issue, we propose a set of four unambiguous computable characteristics (features) of time-varying HIV viral load patterns, along with a novel centroid-based classification algorithm, which we use to classify a population of 1,576 HIV positive clinic patients into one of five different viral load patterns (clusters) often found in the literature: durably suppressed viral load (DSVL), sustained low viral load (SLVL), sustained high viral load (SHVL), high viral load suppression (HVLS), and rebounding viral load (RVL). The centroid algorithm summarizes these clusters in terms of their centroids and radii. We show that this allows new viral load patterns to be assigned pattern membership based on the distance from the centroid relative to its radius, which we term radial normalization classification. This method has the benefit of providing an objective and quantitative method to assign viral load pattern membership with a concise and interpretable model that aids clinical decision making. This method also facilitates meta-analyses by providing computably distinct HIV categories. Finally we propose that this novel centroid algorithm could also be useful in the areas of cluster comparison for outcomes research and data reduction in machine learning.

Pub.: 25 Apr '18, Pinned: 10 May '18

RNA and DNA Sanger sequencing versus next-generation sequencing for HIV-1 drug resistance testing in treatment-naive patients.

Abstract: Sanger sequencing of plasma RNA is the standard method for HIV-1 drug resistance testing in treatment-naive patients, but is limited by the non-detection of resistance-associated mutations (RAMs) with prevalence below approximately 20%.We compared RNA and DNA Sanger sequencing (RSS and DSS) with RNA next-generation sequencing (NGS) for RAM detection in HIV-1 reverse transcriptase (RT), protease (PR) and integrase (IN) genes.Sanger sequencing was performed on RNA and DNA, following the recommendations of the French Agency for AIDS Research (ANRS). NGS was performed on RNA using the HIV-1 Drug Resistance Assay, v. 3.0 (Roche) on the 454 GS Junior sequencer. The IAS-USA list was used to identify RAMs. ANRS, Rega and Stanford algorithms were used for drug resistance interpretation.The study included 48 ART-naive patients. The number of patients with at least one major RAM was 3, 3, 4 and 8 when using RSS, DSS, NGS 20% and NGS 5%, respectively. Numerous minor mutations were detected in patients, especially in the protease gene. None of the methods detected any major mutation in the integrase gene. Overall, the mutation detection rate was similar between RSS and DSS, and higher with NGS 20%. Differences in drug resistance interpretation were found between algorithms. No impact of the minority RAMs detected by NGS was found on the short-term treatment outcome.DSS does not clearly improve the detection of RAMs in ART-naive patients, as compared with RSS. NGS allows detection of additional minority RAMs; however, their clinical relevance requires further investigation.

Pub.: 02 Nov '17, Pinned: 10 May '18

Machine Learning to Improve the Effectiveness of ANRS in Predicting HIV Drug Resistance.

Abstract: Human immunodeficiency virus infection and acquired immune deficiency syndrome (HIV/AIDS) is one of the major burdens of disease in developing countries, and the standard-of-care treatment includes prescribing antiretroviral drugs. However, antiretroviral drug resistance is inevitable due to selective pressure associated with the high mutation rate of HIV. Determining antiretroviral resistance can be done by phenotypic laboratory tests or by computer-based interpretation algorithms. Computer-based algorithms have been shown to have many advantages over laboratory tests. The ANRS (Agence Nationale de Recherches sur le SIDA) is regarded as a gold standard in interpreting HIV drug resistance using mutations in genomes. The aim of this study was to improve the prediction of the ANRS gold standard in predicting HIV drug resistance.A genome sequence and HIV drug resistance measures were obtained from the Stanford HIV database (http://hivdb.stanford.edu/). Feature selection was used to determine the most important mutations associated with resistance prediction. These mutations were added to the ANRS rules, and the difference in the prediction ability was measured.This study uncovered important mutations that were not associated with the original ANRS rules. On average, the ANRS algorithm was improved by 79% ± 6.6%. The positive predictive value improved by 28%, and the negative predicative value improved by 10%.The study shows that there is a significant improvement in the prediction ability of ANRS gold standard.

Pub.: 29 Nov '17, Pinned: 10 May '18

Appearance of drug resistance mutations among the dominant HIV-1 subtype, CRF01_AE in Maumere, Indonesia.

Abstract: Human Immunodeficiency Virus (HIV) is still a major health issue in Indonesia. In recent years, the appearance of drug resistance-associated mutations has reduced the effectiveness of antiretroviral therapy (ART). We conducted genotypic studies, including the detection of drug resistance-associated mutations (from first-line regimen drugs), on HIV-1 genes derived from infected individuals in Maumere, West Nusa Tenggara. Maumere, a transit city in West Nusa Tenggara, which has a high HIV-1 transmission rate. We collected 60 peripheral blood samples from 53 ART-experienced and 7 ART-naive individuals at TC Hillers Hospital, Maumere between 2014 and 2015. The amplification and a sequencing analysis of pol genes encoding protease (the PR gene) and reverse transcriptase (the RT gene) as well as the viral env and gag genes were performed. HIV-1 subtyping and the detection of drug resistance-associated mutations were then conducted. Among 60 samples, 46 PR, 31 RT, 30 env, and 20 gag genes were successfully sequenced. The dominant HIV-1 subtype circulating in Maumere was CRF01_AE. Subtype B and recombinant viruses containing gene fragments of CRF01_AE, subtypes A, B, C, and/or G were also identified as minor populations. The major drug resistance-associated mutations, M184V, K103N, Y188L, and M230I, were found in the RT genes. However, no major drug resistance-associated mutations were detected in the PR genes. CRF01_AE was the major HIV-1 subtype prevalent in Maumere. The appearance of drug resistance-associated mutations found in the present study supports the necessity of monitoring the effectiveness of ART in Maumere. Copyright© Bentham Science Publishers; For any queries, please email at epub@benthamscience.org.

Pub.: 08 May '18, Pinned: 10 May '18