Patrícia R. Oliveira

Applying machine learning techniques for ADME-Tox prediction: a review

Introduction: Pharmacokinetics involves the study of absorption, distribution, metabolism, excretion and toxicity of xenobiotics (ADME-Tox). In this sense, the ADME-Tox profile of a bioactive compound can impact its efficacy and safety. Moreover, efficacy and safety were considered some of the major causes of clinical failures in the development of new chemical entities. In this context, machine learning (ML) techniques have been often used in ADME-Tox studies due to the existence of compounds with known pharmacokinetic properties available for generating predictive models. Areas covered: This review examines the growth in the use of some ML techniques in ADME-Tox studies, in particular supervised and unsupervised techniques. Also, some critical points (e.g., size of the data set and type of output variable) must be considered during the generation of models that relate ADME-Tox properties and biological activity. Expert opinion: ML techniques have been successfully employed in pharmacokinetic studies, helping the complex process of designing new drug candidates from the use of reliable ML models. An application of this procedure would be the prediction of ADME-Tox properties from studies of quantitative structure–activity relationships or the discovery of new compounds from a virtual screening using filters based on results obtained from ML techniques.

Improvements on ICA mixture models for image pre-processing and segmentation

Today several different unsupervised classification algorithms are commonly used to cluster similar patterns in a data set based only on its statistical properties. Specially in image data applications, self-organizing methods for unsupervised classification have been successfully applied for clustering pixels or group of pixels in order to perform segmentation tasks. The first important contribution of this paper refers to the development of a self-organizing method for data classification, named Enhanced Independent Component Analysis Mixture Model (EICAMM), which was built by proposing some modifications in the Independent Component Analysis Mixture Model (ICAMM). Such improvements were proposed by considering some of the model limitations as well as by analyzing how it should be improved in order to become more efficient. Moreover, a pre-processing methodology was also proposed, which is based on combining the Sparse Code Shrinkage (SCS) for image denoising and the Sobel edge detector. In the experiments of this work, the EICAMM and other self-organizing models were applied for segmenting images in their original and pre-processed versions. A comparative analysis showed satisfactory and competitive image segmentation results obtained by the proposals presented herein.

Automatic segmentation of breast masses using enhanced ICA mixture model

Abstract Image segmentation is considered, among all the stages of image processing, the most critical stage of data processing, because a good classification is dependent on the features extracted from the segmented images. In this work, we are proposing to use the technique called Enhanced ICA Mixture Model (EICAMM) for automatic segmentation of breast masses, aiming to comparing it to other segmentation methods known for segmentation of medical images such as Watershed, Self-Organizing Map (SOM), K-means and Fuzzy C-means techniques. For the analysis of the results, it was used Jaccard similarity measure for comparing the result obtained by the segmentation techniques with that one obtained by an expert opinion. All images considered in this work were segmented and then analyzed by us to improve the segmentation performed by an expert and to detect lesion shape for further classification. These models have been applied for the segmentation of suspicious masses in digital mammographic images, including images of dense breasts. The obtained results show a good performance of EICAMM that was the unique technique able to detect masses in dense breast interest region in preprocessed images and in original images. In this way, EICAMM could be considered as a good alternative approach to be applied for breast masses classification.

A comparision between PCA neural networks and the JPEG standard for performing image compression

Principal component analysis (PCA), also called Karhunen-Loeve transform, is a statistical method for multivariate data analysis that can be used in particular to reduce the data set being considered. There are two approaches for performing PCA. The first utilizes the classical statistical method and the other, artificial neural networks. In this paper, neural networks that performing PCA are presented and used to realize tomographic image compression. The results obtained are compared to that obtained by using JPEG compression standard technique and show the usefulness of neural networks for performing image compression.

Transfer and Multi-task Learning in QSAR Modeling: Advances and Challenges

Medicinal chemistry projects involve some steps aiming to develop a new drug, such as the analysis of biological targets related to a given disease, the discovery and the development of drug candidates for these targets, performing parallel biological tests to validate the drug effectiveness and side effects. Approaches as quantitative study of activity-structure relationships (QSAR) involve the construction of predictive models that relate a set of descriptors of a chemical compound series and its biological activities with respect to one or more targets in the human body. Datasets used to perform QSAR analyses are generally characterized by a small number of samples and this makes them more complex to build accurate predictive models. In this context, transfer and multi-task learning techniques are very suitable since they take information from other QSAR models to the same biological target, reducing efforts and costs for generating new chemical compounds. Therefore, this review will present the main features of transfer and multi-task learning studies, as well as some applications and its potentiality in drug design projects.

Portuguese pronoun resolution: resources and evaluation

Despite being one of the most widely-spoken languages in the world, Portuguese remains a relatively resource-poor language, for which only in recently years NLP tools such as parsers, taggers and (fairly) large corpora have become available. In this work we describe the task of pronominal co-reference annotation and resolution in Portuguese texts, in which we take advantage of information provided by a tagged corpus and a simple annotation tool that has been developed for this purpose. Besides developing some of these basic resources from scratch, our ultimate goal is to investigate the multilingual resolution of Portuguese personal pronouns to improve the accuracy of their translations to both Spanish and English in an underlying MT project.

Enhanced ICA Mixture Model for Unsupervised Classification

The ICA mixture model was originally proposed to perform unsupervised classification of data modelled as a mixture of classes described by linear combinations of independent, non-Gaussian densities. Since the original learning algorithm is based on a gradient optimization technique, it was noted that its performance is affected by some known limitations associated with this kind of approach. In this paper, improvements based on implementation and modelling aspects are incorporated to ICA mixture model aiming to achieve better classification results. Comparative experimental results obtained by the enhanced method and the original one are presented to show that the proposed modifications can significantly improve the classification performance considering random generated data and the well-known iris flower data set.

Enhanced ICA mixture model for image segmentation

The ICA mixture model has been proposed to perform unsupervised classification of data modelled as a mixture of classes described by linear combinations ql independent, non-Gaussian densities. Since the original learning algorithm is based on a gradient optimization technique, it was noted that its performance is affected by some known limitations associated with this kind of approach. In this paper, improvements based on implementation and modelling aspects are incorporated to ICA mixture model aiming to apply it for image segmentation. Comparative experimental results obtained by the enhanced method and the original one are presented to show that the proposed modifications can significantly improve the classification and segmentation performance considering random generated data and some image data of public domain.

Techniques for image compression: a comparative analysis

Some techniques for image compression are investigated in this article. The first one is the well known JPEG that is the most widely used technique for image compression. The second is principal component analysis (PCA), also called Karhunen-Loeve transform, that is a statistical method applied for multivariate data analysis and feature extraction. In the latter, two approaches are being considered. The first approach uses the classical statistical method and the other one is based on artificial neural networks. In a comparative study, the results obtained by PCA neural network for compressing medical images are analyzed together with those obtained by using the classical statistical method and JPEG compression standard technique.

A molecular modeling study of combretastatin-like chalcones as anticancer agents using PLS, ANN and consensus models

Combretastatin-like chalcones are promising anticancer compounds that inhibit the mitotic process through interactions with β-tubulin. A detailed study of these compounds can contribute for the rational drug design of new structures aiming at compounds with high biological activity. For this purpose, we have studied 87 combretastatin-like chalcones and proposed multivariate models based on partial least squares (PLS), artificial neural network consensus model (ANN-CM), and general consensus model (GCM). The proposed models have showed good predictive ability with r2test = 0.812 and MSE (test set) = 0.327 for the PLS model, r2test = 0.829 and MSE (test set) = 0.286 for the ANN-CM, and r2test = 0.822 and and MSE (test set) = 0.302 for the GCM. The selected molecular and electronic descriptors (RDF045e, RTv, RDF155u, RDF035m, SP02, PI, UNIP and EHOMO-3) represent molecular features of the compounds that can be associated to the biological activity and can be employed to help the design of new bioactive ligands with improved biological activity.