Predicting HIV-1 Protease Cleavage Sites With Positive-Unlabeled Learning

Abstract

Understanding the substrate specificity of HIV-1 protease plays an essential role in the prevention of HIV infection. A variety of computational models have thus been developed to predict substrate sites that are cleaved by HIV-1 protease, but most of them normally follow a supervised learning scheme to build classifiers by considering experimentally verified cleavable sites as positive samples and unknown sites as negative samples. However, certain noisy can be contained in the negative set, as false negative samples are possibly existed. Hence, the performance of the classifiers is not as accurate as they could be due to the biased prediction results. In this work, unknown substrate sites are regarded as unlabeled samples instead of negative ones. We propose a novel positive-unlabeled learning algorithm, namely PU-HIV, for an effective prediction of HIV-1 protease cleavage sites. Features used by PU-HIV are encoded from different perspectives of substrate sequences, including amino acid identities, coevolutionary patterns and chemical properties. By adjusting the weights of errors generated by positive and unlabeled samples, a biased support vector machine classifier can be built to complete the prediction task. In comparison with state-of-the-art prediction models, benchmarking experiments using cross-validation and independent tests demonstrated the superior performance of PU-HIV in terms of AUC, PR-AUC, and F-measure. Thus, with PU-HIV, it is possible to identify previously unknown, but physiologically existed substrate sites that are able to be cleaved by HIV-1 protease, thus providing valuable insights into designing novel HIV-1 protease inhibitors for HIV treatment.

Prediction of HIV-1 protease cleavage site from octapeptide sequence information using selected classifiers and hybrid descriptors

“…Furthermore, using diverse crossvalidated performance metrics is considered good practice and can objectively reveal the true performance of a model rather than depending on just one metric that could be biased towards a subset of the dataset. This approach reduces the risk of overfitting [6,18,35]. Our models performance was also evaluated on an independent testing set which was not previously exposed to the models to give a true account of the models predictive strength.…”

“…As a result, feature selection is critical in classification tasks [35]. The number and type of peptide feature descriptors selected determine to a great extent the performance of a model [6]. Amino acids are the building blocks of peptides and proteins.…”

Effectively predicting HIV-1 protease cleavage sites by using an ensemble learning approach

“…As a popular metric for binary classification problems, F-measure indicates the harmonic mean of Precision and Recall. The details of computing F-measure can be found in [ 31 ].…”

“…Combining the knowledge from experimental studies, a multitask learning model is developed recently based on multi-kernel [ 30 ], and it utilizes the dependencies among various related tasks to build a stronger predictive model for HIV-1 protease cleavage sites prediction. Since certain noisy can be contained by mislabeling cleavable octamers as negative instances, PU-HIV [ 31 ] considers unknown substrate sites as unlabeled samples, and makes use of positive-unlabeled learning to effectively predict HIV-1 protease cleavage sites.

Fig.

…”

“…With the development of machine learning techniques in bioinformatics [11], a variety of machine learning-based methods have been developed to effectively predict the existence of HIV-1 protease cleavage sites in the substrates [12][13][14][15][16][17][18][19][20][21][22][23][24][25][26][27][28][29][30][31]. They usually regard the prediction problem as a typical binary classification task, which is then achieved with a two-step procedure.…”