Machine Learning for Prediction of HIV Drug Resistance: A Review

Bonet, Isis

doi:10.2174/1574893610666151008011731

Cited by 11 publications

(7 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We built the regression models for each drug separately, taking each polymorphic protein position as a predictor variable and the drug resistance value as the target variable. Since the distributions of resistances are highly skewed we used the log-transformed values, as recommended in [5]. Redundant viruses obtained from the same patient were removed to minimize bias.…”

Section: Methodsmentioning

confidence: 99%

“…This can be either gene sequence or the translated protein sequence; this latter approach eliminates the noise of synonymous mutations. In any case, as genome sequencing is cheaper, faster and more widely available than performing an in vitro drug susceptibility test, much effort has been invested in developing algorithms that predict the drug resistance from the virus sequence [5].…”

Section: Introductionmentioning

confidence: 99%

“…The rule-based algorithms continue to be used to this day because of their interpretability. Some publicly available examples are the Stanford HIVdb, Rega or ANRS softwares [5]. However, the aforementioned high mutation rate of the HIV, which favors the emergence of large numbers of new resistance mutations and complex mutational patterns, makes the rule-based approach suboptimal.…”

Section: Introductionmentioning

confidence: 99%

“…However, most machine learning algorithms are designed to cope with numeric data (DT and RF being exceptions), thus obliging to perform some kind of pre-processing. A typical approach is to recode each position into m or m − 1 “dummy variables”, which can take the values 0 or 1 [5]. Usually, m is the number of all possible alleles that can be potentially found in a position (i.e., m = 20 in protein sequences).…”

Section: Introductionmentioning

confidence: 99%

“…A very different approach is found in [14], where each amino acid was codified as an integer ranging 1–22 (the 20 canonical amino acids plus two extra characters B and Z). Other encodings have been used with HIV sequence data, like amino acid composition frequencies, reduced amino acid alphabets or physicochemical properties [5, 16, 20].…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

HIV drug resistance prediction with weighted categorical kernel functions

2019

View full text Add to dashboard Cite

Background Antiretroviral drugs are a very effective therapy against HIV infection. However, the high mutation rate of HIV permits the emergence of variants that can be resistant to the drug treatment. Predicting drug resistance to previously unobserved variants is therefore very important for an optimum medical treatment. In this paper, we propose the use of weighted categorical kernel functions to predict drug resistance from virus sequence data. These kernel functions are very simple to implement and are able to take into account HIV data particularities, such as allele mixtures, and to weigh the different importance of each protein residue, as it is known that not all positions contribute equally to the resistance. Results We analyzed 21 drugs of four classes: protease inhibitors (PI), integrase inhibitors (INI), nucleoside reverse transcriptase inhibitors (NRTI) and non-nucleoside reverse transcriptase inhibitors (NNRTI). We compared two categorical kernel functions, Overlap and Jaccard, against two well-known noncategorical kernel functions (Linear and RBF) and Random Forest (RF). Weighted versions of these kernels were also considered, where the weights were obtained from the RF decrease in node impurity. The Jaccard kernel was the best method, either in its weighted or unweighted form, for 20 out of the 21 drugs. Conclusions Results show that kernels that take into account both the categorical nature of the data and the presence of mixtures consistently result in the best prediction model. The advantage of including weights depended on the protein targeted by the drug. In the case of reverse transcriptase, weights based in the relative importance of each position clearly increased the prediction performance, while the improvement in the protease was much smaller. This seems to be related to the distribution of weights, as measured by the Gini index. All methods described, together with documentation and examples, are freely available at https://bitbucket.org/elies_ramon/catkern . Electronic supplementary material The online version of this article (10.1186/s12859-019-2991-2) contains supplementary material, which is available to authorized users.

show abstract

Section: Methodsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%