Visual Characterization of Misclassified Class C GPCRs through Manifold-based Machine Learning Methods

Cárdenas, Martha Ivón; Vellido, Alfredo; König, Caroline; Alquézar, René; Giraldo, Jesús

doi:10.18547/gcb.2015.vol1.iss1.e19

Cited by 4 publications

(3 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…These results were further confirmed from the viewpoint of visualization-oriented fully unsupervised machine learning methods (that is, methods that attempted sequence discrimination into subtypes without knowledge of sequence-to-subtype assignment). Results clearly indicated that the subtypes shown to be worse discriminated by supervised classifiers were also those shown to heavily overlap in unsupervised visualization models from different unaligned sequence data transformations 31 .…”

Section: Datamentioning

confidence: 99%

Using machine learning tools for protein database biocuration assistance

König

Shaim

Vellido

et al. 2018

Sci Rep

Self Cite

View full text Add to dashboard Cite

Biocuration in the omics sciences has become paramount, as research in these fields rapidly evolves towards increasingly data-dependent models. As a result, the management of web-accessible publicly-available databases becomes a central task in biological knowledge dissemination. One relevant challenge for biocurators is the unambiguous identification of biological entities. In this study, we illustrate the adequacy of machine learning methods as biocuration assistance tools using a publicly available protein database as an example. This database contains information on G Protein-Coupled Receptors (GPCRs), which are part of eukaryotic cell membranes and relevant in cell communication as well as major drug targets in pharmacology. These receptors are characterized according to subtype labels. Previous analysis of this database provided evidence that some of the receptor sequences could be affected by a case of label noise, as they appeared to be too consistently misclassified by machine learning methods. Here, we extend our analysis to recent and quite substantially modified new versions of the database and reveal their now extremely accurate labeling using several machine learning models and different transformations of the unaligned sequences. These findings support the adequacy of our proposed method to identify problematic labeling cases as a tool for database biocuration.

show abstract

Section: Datamentioning

confidence: 99%

Using machine learning tools for protein database biocuration assistance

König

Shaim

Vellido

et al. 2018

Sci Rep

Self Cite

View full text Add to dashboard Cite

show abstract

“…Results clearly indicated that the subtypes shown to be worse discriminated by supervised classifiers were also those shown to heavily overlap in unsupervised visualization models from different unaligned sequence data transformations [6].…”

Section: Previous Research On Gpcr Class C From a Data Curation Persp...mentioning

confidence: 99%

“…[4] misclassification analysis [5,6] Tracking the evolution of class C GPCR chapter 6 [7] database Topological sequence segments discriminate chapter 7 [8,9] betw. class C GPCR subtypes Feature selection for the identification chapter 8 [10,11] of subtype-discriminating n-grams.…”

Section: Topicmentioning

confidence: 99%

Analysis of class C G-protein coupled receptors using supervised classification methods

König

View full text Add to dashboard Cite

G protein-coupled receptors (GPCRs) are cell membrane proteins with a key role in regulating the function of cells. This is the result of their ability to transmit extracellular signals, which makes them relevant for pharmacology and has led, over the last decade, to active research in the field of proteomics. The current thesis specifically targets class C of GPCRs, which are relevant in therapies for various central nervous system disorders, such as Alzheimer’s disease, anxiety, Parkinson’s disease and schizophrenia. The investigation of protein functionality often relies on the knowledge of crystal three dimensional (3-D) structures, which determine the receptor’s ability for ligand binding responsible for the activation of certain functionalities in the protein. The structural information is therefore paramount, but it is not always known or easily unravelled, which is the case of eukaryotic cell membrane proteins such as GPCRs. In the face of the lack of information about the 3-D structure, research is often bound to the analysis of the primary amino acid sequences of the proteins, which are commonly known and available from curated databases. Much research on sequence analysis has focused on the quantitative analysis of their aligned versions, although, recently, alternative approaches using machine learning techniques for the analysis of alignment-free sequences have been proposed. In this thesis, we focus on the differentiation of class C GPCRs into functional and structural related subgroups based on the alignment-free analysis of their sequences using supervised classification models. In the first part of the thesis, the main topic is the construction of supervised classification models for unaligned protein sequences based on physicochemical transformations and n-gram representations of their amino acid sequences. These models are useful to assess the internal data quality of the externally labeled dataset and to manage the label noise problem from a data curation perspective. In its second part, the thesis focuses on the analysis of the sequences to discover subtype- and region-speci¿c sequence motifs. For that, we carry out a systematic analysis of the topological sequence segments with supervised classification models and evaluate the subtype discrimination capability of each region. In addition, we apply different types of feature selection techniques to the n-gram representation of the amino acid sequence segments to find subtype and region specific motifs. Finally, we compare the findings of this motif search with the partially known 3D crystallographic structures of class C GPCRs. Los receptores acoplados a proteínas G (GPCRs) son proteínas de la membrana celular con un papel clave para la regulación del funcionamiento de una célula. Esto es consecuencia de su capacidad de transmisión de señales extracelulares, lo que les hace relevante en la farmacología y que ha llevado a investigaciones activas en la última década en el área de la proteómica. Esta tesis se centra específicamente en la clase C de GPCRs, que son relevante para terapias de varios trastornos del sistema nervioso central, como la enfermedad de Alzheimer, ansiedad, enfermedad de Parkinson y esquizofrenia. La investigación de la funcionalidad de proteínas muchas veces se basa en el conocimiento de la estructura cristalina tridimensional (3-D), que determina la capacidad del receptor para la unión con ligandos, que son responsables para la activación de ciertas funcionalidades en la proteína. El análisis de secuencias de amino ácidos se ha centrado en muchas investigaciones en el análisis cuantitativo de las versiones alineados de las secuencias, aunque, recientemente, se han propuesto métodos alternativos usando métodos de aprendizaje automático aplicados a las versiones no-alineadas de las secuencias. En esta tesis, nos centramos en la diferenciación de los GPCRs de la clase C en subgrupos funcionales y estructurales basado en el análisis de las secuencias no-alineadas utilizando modelos de clasificación supervisados. Estos modelos son útiles para evaluar la calidad interna de los datos a partir del conjunto de datos etiquetados externamente y para gestionar el problema del 'ruido de datos' desde la perspectiva de la curación de datos. En su segunda parte, la tesis enfoca el análisis de las secuencias para descubrir motivos de secuencias específicos a nivel de subtipo o región. Para eso, llevamos a cabo un análisis sistemático de los segmentos topológicos de la secuencia con modelos supervisados de clasificación y evaluamos la capacidad de discriminar entre subtipos de cada región. Adicionalmente, aplicamos diferentes tipos de técnicas de selección de atributos a las representaciones mediante n-gramas de los segmentos de secuencias de amino ácidos para encontrar motivos específicos a nivel de subtipo y región. Finalmente, comparamos los descubrimientos de la búsqueda de motivos con las estructuras cristalinas parcialmente conocidas para la clase C de GPCRs.

show abstract