RESUMEN
The use of non-invasive liquid biopsy-based cell-free DNA (cfDNA) analysis is an emerging method of cancer detection and intervention. Different analytical methodologies are used to investigate cfDNA characteristics, resulting in costly and long analysis processes needed for combining different data. This study investigates the possibility of using cfDNA data converted for methylation analysis for combining the cfDNA fragment size with copy number variation (CNV) in the context of early colorectal cancer detection. Specifically, we focused on comparing enzymatically and bisulfite-converted data for evaluating cfDNA fragments belonging to chromosome 18. Chromosome 18 is often reported to be deleted in colorectal cancer. We used counts of short and medium cfDNA fragments of chromosome 18 and trained a linear model (LDA) on a set of 2959 regions to predict early-stage (I-IIA) colorectal cancer on an independent test set. In total, 87.5% sensitivity and 92% specificity were obtained on the enzymatically converted libraries. Repeating the same workflow on bisulfite-converted data yielded lower accuracy results with 58.3% sensitivity, implying that enzymatic conversion preserves the cancer fragmentation footprint in whole genome data better than bisulfite conversion. These results could serve as a promising new avenue for the early detection of colorectal cancer using fragmentation and methylation approaches on the same datasets.
Asunto(s)
Ácidos Nucleicos Libres de Células , Neoplasias Colorrectales , Sulfitos , Humanos , Ácidos Nucleicos Libres de Células/genética , Variaciones en el Número de Copia de ADN/genética , Metilación de ADN/genética , Neoplasias Colorrectales/diagnóstico , Neoplasias Colorrectales/genética , Biomarcadores de Tumor/genéticaRESUMEN
BACKGROUND: In this study, we evaluated the lipidome alterations caused by type 1 diabetes (T1D) and type 2 diabetes (T2D), by determining lipids significantly associated with diabetes overall and in both sexes, and lipids associated with the glycaemic state. METHODS: An untargeted lipidomic analysis was performed to measure the lipid profiles of 360 subjects (91 T1D, 91 T2D, 74 with prediabetes and 104 controls (CT)) without cardiovascular and/or chronic kidney disease. Ultra-high performance liquid chromatography-electrospray ionization mass spectrometry (UHPLC-ESI-MS) was conducted in two ion modes (positive and negative). We used multiple linear regression models to (1) assess the association between each lipid feature and each condition, (2) determine sex-specific differences related to diabetes, and (3) identify lipids associated with the glycaemic state by considering the prediabetes stage. The models were adjusted by sex, age, hypertension, dyslipidaemia, body mass index, glucose, smoking, systolic blood pressure, triglycerides, HDL cholesterol, LDL cholesterol, alternate Mediterranean diet score (aMED) and estimated glomerular filtration rate (eGFR); diabetes duration and glycated haemoglobin (HbA1c) were also included in the comparison between T1D and T2D. RESULTS: A total of 54 unique lipid subspecies from 15 unique lipid classes were annotated. Lysophosphatidylcholines (LPC) and ceramides (Cer) showed opposite effects in subjects with T1D and subjects with T2D, LPCs being mainly up-regulated in T1D and down-regulated in T2D, and Cer being up-regulated in T2D and down-regulated in T1D. Also, Phosphatidylcholines were clearly down-regulated in subjects with T1D. Regarding sex-specific differences, ceramides and phosphatidylcholines exhibited important diabetes-associated differences due to sex. Concerning the glycaemic state, we found a gradual increase of a panel of 1-deoxyceramides from normoglycemia to prediabetes to T2D. CONCLUSIONS: Our findings revealed an extensive disruption of lipid metabolism in both T1D and T2D. Additionally, we found sex-specific lipidome changes associated with diabetes, and lipids associated with the glycaemic state that can be linked to previously described molecular mechanisms in diabetes.
Asunto(s)
Diabetes Mellitus Tipo 1 , Diabetes Mellitus Tipo 2 , Estado Prediabético , Masculino , Femenino , Humanos , Lipidómica , Estado Prediabético/diagnóstico , Estado Prediabético/complicaciones , HDL-Colesterol , Ceramidas , FosfatidilcolinasRESUMEN
OBJECTIVE: This study aims to comprehend the impact of handball practice on sub-elite athletes by investigating transcriptomic changes that occur during a match. The primary focus encompasses a dual objective: firstly, to identify and characterize these transcriptomic alterations, and secondly, to establish correlations between internal factors (gene expression), and external loads measured through Electronic Performance and Tracking Systems (EPTS variables). Ultimately, this comprehensive analysis seeks to evaluate both acute and chronic responses to exercise within the context of handball training. METHODS: The study included sixteen elite male athletes from the FC Barcelona handball second team. Blood samples were extracted at three different time points: before the match at baseline levels (T1), immediately upon completion (T2), and 24 hours after completion (T3). Differential gene expression, Gene Ontology Term and Kyoto Encyclopedia of Genes and Genomes pathway enrichment analyses were conducted in two comparisons: Comparison 1 (T1 vs T2) and Comparison 2 (T1 vs T3). Further, the correlation between gene expression levels and training variables (external load) was conducted. RESULTS: In T1 vs T2, 3717 of the 14632 genes detected were differentially expressed (adjusted p-value < 0.05), and enrichment of terms related to the immune system, mitochondria, and metabolic processes was found. Further, significant linear correlations were obtained between High-Speed running (HSR) and high-intensity variables such as acceleration ACC and deceleration DEC values with amino acids, and inflammatory and oxidative environment-related pathways, both in chronic and acute response. CONCLUSIONS: This research highlights the effects of external workload on elite athletes during a handball match and throughout the season. The study identifies deregulation in the immune system, mitochondrial functions, and various metabolic pathways during the match. Additionally, it establishes correlations between the external load and pathways associated with amino acids, inflammation, oxidative environment, and regulation. These findings offer insights into the immediate and chronic responses of athletes to physical effort.
Asunto(s)
Rendimiento Atlético , Carrera , Humanos , Masculino , Rendimiento Atlético/fisiología , Carrera/fisiología , Atletas , Perfilación de la Expresión Génica , AminoácidosRESUMEN
Lifestyle interventions have positive neuroprotective effects in aging. However, there are still open questions about how changes in resting-state functional connectivity (rsFC) contribute to cognitive improvements. The Projecte Moviment is a 12-week randomized controlled trial of a multimodal data acquisition protocol that investigated the effects of aerobic exercise (AE), computerized cognitive training (CCT), and their combination (COMB). An initial list of 109 participants was recruited from which a total of 82 participants (62% female; age = 58.38 ± 5.47) finished the intervention with a level of adherence > 80%. Only in the COMB group, we revealed an extended network of 33 connections that involved an increased and decreased rsFC within and between the aDMN/pDMN and a reduced rsFC between the bilateral supplementary motor areas and the right thalamus. No global and especially local rsFC changes due to any intervention mediated the cognitive benefits detected in the AE and COMB groups. Projecte Moviment provides evidence of the clinical relevance of lifestyle interventions and the potential benefits when combining them.
Asunto(s)
Encéfalo , Entrenamiento Cognitivo , Humanos , Femenino , Persona de Mediana Edad , Masculino , Ejercicio Físico , Mapeo Encefálico/métodos , Estado de SaludRESUMEN
Metabolomics is the latest of the omics sciences. It attempts to measure and characterize metabolites-small chemical compounds <1500 Da-on cells, tissue, or biofluids, which are usually products of biological reactions. As metabolic reactions are closer to the phenotype, metabolomics has emerged as an attractive science for various areas of research, including personalized medicine. However, due to the complexity of data obtained and the absence of curated databases for metabolite identification, data processing is the major bottleneck in this area since most technicians lack the required bioinformatics expertise to process datasets in a reliable and fast manner. The aim of this chapter is to describe the available tools for data processing that makes an inexperienced researcher capable of obtaining reliable results without having to undergo through huge parametrization steps.
Asunto(s)
Metabolómica , Bases de Datos Factuales , Espectrometría de Masas/métodos , Metabolómica/métodos , FenotipoRESUMEN
Background: Post-stroke cognitive and emotional complications are frequent in the chronic stages of stroke and have important implications for the functionality and quality of life of those affected and their caregivers. Strategies such as mindfulness meditation, physical exercise (PE), or computerized cognitive training (CCT) may benefit stroke patients by impacting neuroplasticity and brain health. Materials and methods: One hundred and forty-one chronic stroke patients are randomly allocated to receive mindfulness-based stress reduction + CCT (n = 47), multicomponent PE program + CCT (n = 47), or CCT alone (n = 47). Interventions consist of 12-week home-based programs five days per week. Before and after the interventions, we collect data from cognitive, psychological, and physical tests, blood and stool samples, and structural and functional brain scans. Results: The effects of the interventions on cognitive and emotional outcomes will be described in intention-to-treat and per-protocol analyses. We will also explore potential mediators and moderators, such as genetic, molecular, brain, demographic, and clinical factors in our per-protocol sample. Discussion: The MindFit Project is a randomized clinical trial that aims to assess the impact of mindfulness and PE combined with CCT on chronic stroke patients' cognitive and emotional wellbeing. Furthermore, our design takes a multimodal biopsychosocial approach that will generate new knowledge at multiple levels of evidence, from molecular bases to behavioral changes. Clinical trial registration: www.ClinicalTrials.gov, identifier NCT04759950.
RESUMEN
Type 2 diabetes mellitus (T2DM) is a highly heterogeneous chronic disease with different pathophysiological and genetic characteristics affecting its progression, associated complications and response to therapies. The advances in deep learning (DL) techniques and the availability of a large amount of healthcare data allow us to investigate T2DM characteristics and evolution with a completely new approach, studying common disease trajectories rather than cross sectional values. We used an Kernelized-AutoEncoder algorithm to map 5 years of data of 11,028 subjects diagnosed with T2DM in a latent space that embedded similarities and differences between patients in terms of the evolution of the disease. Once we obtained the latent space, we used classical clustering algorithms to create longitudinal clusters representing different evolutions of the diabetic disease. Our unsupervised DL clustering algorithm suggested seven different longitudinal clusters. Different mean ages were observed among the clusters (ranging from 65.3±11.6 to 72.8±9.4). Subjects in clusters B (Hypercholesteraemic) and E (Hypertensive) had shorter diabetes duration (9.2±3.9 and 9.5±3.9 years respectively). Subjects in Cluster G (Metabolic) had the poorest glycaemic control (mean glycated hemoglobin 7.99±1.42%), while cluster E had the best one (mean glycated hemoglobin 7.04±1.11%). Obesity was observed mainly in clusters A (Neuropathic), C (Multiple Complications), F (Retinopathy) and G. A dashboard is available at dm2.b2slab.upc.edu to visualize the different trajectories corresponding to the 7 clusters.
Asunto(s)
Aprendizaje Profundo , Diabetes Mellitus Tipo 2 , Humanos , Diabetes Mellitus Tipo 2/epidemiología , Diabetes Mellitus Tipo 2/diagnóstico , Hemoglobina Glucada/análisis , Estudios Transversales , Análisis por ConglomeradosRESUMEN
Early in the COVID-19 pandemic, the Diamond Princess became the center of the largest outbreak outside the original epicenter in China. This outbreak which left 712 passengers infected and 14 dead, followed by subsequent outbreaks affecting over one-third of the active ships in the cruise industry's global fleet, quickly became a crisis that captured public attention and dominated mainstream news and social media. This study investigates the perception of cruising during these outbreaks by analyzing the tweets on cruising using Natural Language Processing (NLP). The findings show a prevalent negative sentiment in most of the analyzed tweets, while the criticisms directed at the cruise industry were based on perceptions and stereotypes of the industry before the pandemic. The study provides insight into the concerns raised in these conversations and highlights the need for new business models outside the pre-pandemic mass-market model and to genuinely make cruising more environmentally friendly.
RESUMEN
Background: During the COVID-19 pandemic, lockdown strategies have been widely used to contain SARS-CoV-2 virus spread. Children and adolescents are especially vulnerable to suffering psychological effects as result of such measures. In Spain, children were enforced to a strict home lockdown for 42 days during the first wave. Here, we studied the effects of lockdown in children and adolescents through an online questionnaire. Methods: A cross-sectional study was conducted in Spain using an open online survey from July (after the lockdown resulting from the first pandemic wave) to November 2020 (second wave). We included families with children under 16 years-old living in Spain. Parents answered a survey regarding the lockdown effects on their children and were instructed to invite their children from 7 to 16 years-old (mandatory scholar age in Spain) to respond a specific set of questions. Answers were collected through an application programming interface system, and data analysis was performed using R. Results: We included 1,957 families who completed the questionnaires, covering a total of 3,347 children. The specific children's questionnaire was completed by 167 kids (7-11 years-old), and 100 adolescents (12-16 years-old). Children, in general, showed high resilience and capability to adapt to new situations. Sleeping problems were reported in more than half of the children (54%) and adolescents (59%), and these were strongly associated with less time doing sports and spending more than 5 h per day using electronic devices. Parents perceived their children to gain weight (41%), be more irritable and anxious (63%) and sadder (46%). Parents and children differed significantly when evaluating children's sleeping disturbances. Conclusions: Enforced lockdown measures and isolation can have a negative impact on children and adolescent's mental health and well-being. In future waves of the current pandemic, or in the light of potential epidemics of new emerging infections, lockdown measures targeting children, and adolescents should be reconsidered taking into account their infectiousness potential and their age-specific needs, especially to facilitate physical activity and to limit time spent on electronic devices.
Asunto(s)
COVID-19 , Adolescente , Niño , Control de Enfermedades Transmisibles , Estudios Transversales , Humanos , Pandemias , Padres , SARS-CoV-2RESUMEN
Untargeted metabolomics using liquid chromatography coupled to mass spectrometry (LC-MS) allows the detection of thousands of metabolites in biological samples. However, LC-MS data annotation is still considered a major bottleneck in the metabolomics pipeline since only a small fraction of the metabolites present in the sample can be annotated with the required confidence level. Here, we introduce mWISE (metabolomics wise inference of speck entities), an R package for context-based annotation of LC-MS data. The algorithm consists of three main steps aimed at (i) matching mass-to-charge ratio values to the Kyoto Encyclopedia of Genes and Genomes (KEGG) database, (ii) clustering and filtering the potential KEGG candidates, and (iii) building a final prioritized list using diffusion in graphs. The algorithm performance is evaluated with three publicly available studies using both positive and negative ionization modes. We have also compared mWISE to other available annotation algorithms in terms of their performance and computation time. In particular, we explored four different configurations for mWISE, and all four of them outperform xMSannotator (a state-of-the-art annotator) in terms of both performance and computation time. Using a diffusion configuration that combines the biological network obtained from the FELLA R package and raw scores, mWISE shows a sensitivity mean (standard deviation) across data sets of 0.63 (0.07), while xMSannotator achieves a sensitivity of 0.55 (0.19). We have also shown that the chemical structures of the compounds proposed by mWISE are closer to the original compounds than those proposed by xMSannotator. Finally, we explore the diffusion prioritization separately, showing its key role in the annotation process. mWISE is freely available on GitHub (https://github.com/b2slab/mWISE) under a GPL license.
Asunto(s)
Algoritmos , Metabolómica , Cromatografía Liquida , Difusión , Espectrometría de Masas , Programas InformáticosRESUMEN
In silico analysis of biological activity data has become an essential technique in pharmaceutical development. Specifically, the so-called proteochemometric models aim to share information between targets in machine learning ligand-target activity prediction models. However, bioactivity data sets used in proteochemometric modeling are usually imbalanced, which could potentially affect the performance of the models. In this work, we explored the effect of different balancing strategies in deep learning proteochemometric target-compound activity classification models while controlling for the compound series bias through clustering. These strategies were (1) no_resampling, (2) resampling_after_clustering, (3) resampling_before_clustering, and (4) semi_resampling. These schemas were evaluated in kinases, GPCRs, nuclear receptors, and proteases from BindingDB. We observed that the predicted proportion of positives was driven by the actual data balance in the test set. Additionally, it was confirmed that data balance had an impact on the performance estimates of the proteochemometric model. We recommend a combination of data augmentation and clustering in the training set (semi_resampling) to mitigate the data imbalance effect in a realistic scenario. The code of this analysis is publicly available at https://github.com/b2slab/imbalance_pcm_benchmark.
Asunto(s)
Aprendizaje Profundo , Simulación por Computador , Ligandos , Aprendizaje AutomáticoRESUMEN
BACKGROUND: Rare disease communities are spread around the globe and segmented by their condition. Little research has been performed on the majority of rare diseases. Most patients who are affected by a rare disease have no research on their condition because of a lack of knowledge due to absence of common groups in the research community. OBJECTIVE: We aimed to develop a safe and secure community of rare disease patients, without geographic or language barriers, to promote research. METHODS: Cocreation design methodology was applied to build Share4Rare, with consultation and input through workshops from a variety of stakeholders (patients, caregivers, clinicians, and researchers). RESULTS: The workshops allowed us to develop a layered version of the platform based on educating patients and caregivers with publicly accessible information, a secure community for the patients and caregivers, and a research section with the purpose of collecting patient information for analysis, which was the core and final value of the platform. CONCLUSIONS: Rare disease research requires global collaboration in which patients and caregivers have key roles. Collective intelligence methods implemented in digital platforms reduce geographic and language boundaries and involve patients in a unique and universal project. Their contributions are essential to increase the amount of scientific knowledge that experts have on rare diseases. Share4Rare has been designed as a global platform to facilitate the donation of clinical information to foster research that matters to patients with rare conditions. The codesign methods with patients have been essential to create a patient-centric design.
RESUMEN
MOTIVATION: Network diffusion and label propagation are fundamental tools in computational biology, with applications like gene-disease association, protein function prediction and module discovery. More recently, several publications have introduced a permutation analysis after the propagation process, due to concerns that network topology can bias diffusion scores. This opens the question of the statistical properties and the presence of bias of such diffusion processes in each of its applications. In this work, we characterized some common null models behind the permutation analysis and the statistical properties of the diffusion scores. We benchmarked seven diffusion scores on three case studies: synthetic signals on a yeast interactome, simulated differential gene expression on a protein-protein interaction network and prospective gene set prediction on another interaction network. For clarity, all the datasets were based on binary labels, but we also present theoretical results for quantitative labels. RESULTS: Diffusion scores starting from binary labels were affected by the label codification and exhibited a problem-dependent topological bias that could be removed by the statistical normalization. Parametric and non-parametric normalization addressed both points by being codification-independent and by equalizing the bias. We identified and quantified two sources of bias-mean value and variance-that yielded performance differences when normalizing the scores. We provided closed formulae for both and showed how the null covariance is related to the spectral properties of the graph. Despite none of the proposed scores systematically outperformed the others, normalization was preferred when the sought positive labels were not aligned with the bias. We conclude that the decision on bias removal should be problem and data-driven, i.e. based on a quantitative analysis of the bias and its relation to the positive entities. AVAILABILITY: The code is publicly available at https://github.com/b2slab/diffuBench and the data underlying this article are available at https://github.com/b2slab/retroData. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.
Asunto(s)
Biología Computacional , Mapas de Interacción de Proteínas , Difusión , Estudios Prospectivos , Proteínas/genéticaRESUMEN
SUMMARY: High-throughput screening yields vast amounts of biological data which can be highly challenging to interpret. In response, knowledge-driven approaches emerged as possible solutions to analyze large datasets by leveraging prior knowledge of biomolecular interactions represented in the form of biological networks. Nonetheless, given their size and complexity, their manual investigation quickly becomes impractical. Thus, computational approaches, such as diffusion algorithms, are often employed to interpret and contextualize the results of high-throughput experiments. Here, we present MultiPaths, a framework consisting of two independent Python packages for network analysis. While the first package, DiffuPy, comprises numerous commonly used diffusion algorithms applicable to any generic network, the second, DiffuPath, enables the application of these algorithms on multi-layer biological networks. To facilitate its usability, the framework includes a command line interface, reproducible examples and documentation. To demonstrate the framework, we conducted several diffusion experiments on three independent multi-omics datasets over disparate networks generated from pathway databases, thus, highlighting the ability of multi-layer networks to integrate multiple modalities. Finally, the results of these experiments demonstrate how the generation of harmonized networks from disparate databases can improve predictive performance with respect to individual resources. AVAILABILITY AND IMPLEMENTATION: DiffuPy and DiffuPath are publicly available under the Apache License 2.0 at https://github.com/multipaths. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.
RESUMEN
The use of raw amino acid sequences as input for deep learning models for protein functional prediction has gained popularity in recent years. This scheme obliges to manage proteins with different lengths, while deep learning models require same-shape input. To accomplish this, zeros are usually added to each sequence up to a established common length in a process called zero-padding. However, the effect of different padding strategies on model performance and data structure is yet unknown. We propose and implement four novel types of padding the amino acid sequences. Then, we analysed the impact of different ways of padding the amino acid sequences in a hierarchical Enzyme Commission number prediction problem. Results show that padding has an effect on model performance even when there are convolutional layers implied. Contrastingly to most of deep learning works which focus mainly on architectures, this study highlights the relevance of the deemed-of-low-importance process of padding and raises awareness of the need to refine it for better performance. The code of this analysis is publicly available at https://github.com/b2slab/padding_benchmark .
Asunto(s)
Archaea/metabolismo , Proteínas Arqueales/metabolismo , Aprendizaje Profundo , Secuencia de AminoácidosRESUMEN
In-silico identification of potential target genes for disease is an essential aspect of drug target discovery. Recent studies suggest that successful targets can be found through by leveraging genetic, genomic and protein interaction information. Here, we systematically tested the ability of 12 varied algorithms, based on network propagation, to identify genes that have been targeted by any drug, on gene-disease data from 22 common non-cancerous diseases in OpenTargets. We considered two biological networks, six performance metrics and compared two types of input gene-disease association scores. The impact of the design factors in performance was quantified through additive explanatory models. Standard cross-validation led to over-optimistic performance estimates due to the presence of protein complexes. In order to obtain realistic estimates, we introduced two novel protein complex-aware cross-validation schemes. When seeding biological networks with known drug targets, machine learning and diffusion-based methods found around 2-4 true targets within the top 20 suggestions. Seeding the networks with genes associated to disease by genetics decreased performance below 1 true hit on average. The use of a larger network, although noisier, improved overall performance. We conclude that diffusion-based prioritisers and machine learning applied to diffusion-based features are suited for drug discovery in practice and improve over simpler neighbour-voting methods. We also demonstrate the large impact of choosing an adequate validation strategy and the definition of seed disease genes.
Asunto(s)
Biología Computacional/métodos , Simulación por Computador , Descubrimiento de Drogas/métodos , Algoritmos , Benchmarking , Bases de Datos Genéticas , Enfermedad/genética , Humanos , Aprendizaje AutomáticoRESUMEN
BACKGROUND: Long-term trajectories of left ventricular ejection fraction (LVEF) in heart failure (HF) patients with preserved EF (HFpEF) remain unclear. Our objective was to assess long-term longitudinal trajectories in consecutive HFpEF patients and the prognostic impact of LVEF dynamic changes over time. METHODS AND RESULTS: Consecutive ambulatory HFpEF patients admitted to a multidisciplinary HF Unit were prospectively evaluated by 2-dimensional echocardiography at baseline and at 1, 3, 5, 7, 9, and 11 years of follow-up. Exclusion criteria were patients having a previous known LVEF <50%, patients undergoing only 1 echocardiogram study, and those with a diagnosis of dilated, noncompaction, alcoholic, or toxic cardiomyopathy. One hundred twenty-six patients (age, 71±13 years; 63% women) were included. The main pathogeneses were valvular disease (36%) and hypertension (28%). Atrial fibrillation was present in 67 patients (53%). The mean number of echocardiographies performed was 3±1.2 per patient. Locally weighted error sum of squares curves showed a smooth decrease of LVEF during the 11-year follow-up that was statistically significant in linear mixed-effects modeling ( P=0.01). Ischemic patients showed a higher decrease than nonischemics. The great majority (88.9%) of patients remained in the HFpEF category during follow-up; 9.5% evolved toward HF with midrange LVEF, and only 1.6% dropped to HF with reduced LVEF. No significant relationship was found between LVEF dynamics in the immediate preceding period and mortality. CONCLUSIONS: LVEF remained ≥50% in the majority of patients with HFpEF for ≤11 years. Only 1.6% of patients evolved to HF with reduced LVEF. Dynamic LVEF changes were not associated with mortality.
Asunto(s)
Insuficiencia Cardíaca/fisiopatología , Volumen Sistólico/fisiología , Anciano , Anciano de 80 o más Años , Femenino , Humanos , Estudios Longitudinales , Masculino , Persona de Mediana Edad , Fenotipo , Pronóstico , Sobrevivientes , Factores de TiempoRESUMEN
Binding prediction between targets and drug-like compounds through deep neural networks has generated promising results in recent years, outperforming traditional machine learning-based methods. However, the generalization capability of these classification models is still an issue to be addressed. In this work, we explored how different cross-validation strategies applied to data from different molecular databases affect to the performance of binding prediction proteochemometrics models. These strategies are (1) random splitting, (2) splitting based on K-means clustering (both of actives and inactives), (3) splitting based on source database, and (4) splitting based both in the clustering and in the source database. These schemas are applied to a deep learning proteochemometrics model and to a simple logistic regression model to be used as baseline. Additionally, two different ways of describing molecules in the model are tested: (1) by their SMILES and (2) by three fingerprints. The classification performance of our deep learning-based proteochemometrics model is comparable to the state of the art. Our results show that the lack of generalization of these models is due to a bias in public molecular databases and that a restrictive cross-validation schema based on compound clustering leads to worse but more robust and credible results. Our results also show better performance when representing molecules by their fingerprints.
Asunto(s)
Aprendizaje Profundo , Informática/métodos , Descubrimiento de Drogas , Relación Estructura-Actividad Cuantitativa , Reproducibilidad de los ResultadosRESUMEN
SUMMARY: Multiview datasets are the norm in bioinformatics, often under the label multi-omics. Multiview data are gathered from several experiments, measurements or feature sets available for the same subjects. Recent studies in pattern recognition have shown the advantage of using multiview methods of clustering and dimensionality reduction; however, none of these methods are readily available to the extent of our knowledge. Multiview extensions of four well-known pattern recognition methods are proposed here. Three multiview dimensionality reduction methods: multiview t-distributed stochastic neighbour embedding, multiview multidimensional scaling and multiview minimum curvilinearity embedding, as well as a multiview spectral clustering method. Often they produce better results than their single-view counterparts, tested here on four multiview datasets. AVAILABILITY AND IMPLEMENTATION: R package at the B2SLab site: http://b2slab.upc.edu/software-and-tutorials/ and Python package: https://pypi.python.org/pypi/multiview. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.
Asunto(s)
Programas Informáticos , Análisis por ConglomeradosRESUMEN
BACKGROUND: Pathway enrichment techniques are useful for understanding experimental metabolomics data. Their purpose is to give context to the affected metabolites in terms of the prior knowledge contained in metabolic pathways. However, the interpretation of a prioritized pathway list is still challenging, as pathways show overlap and cross talk effects. RESULTS: We introduce FELLA, an R package to perform a network-based enrichment of a list of affected metabolites. FELLA builds a hierarchical representation of an organism biochemistry from the Kyoto Encyclopedia of Genes and Genomes (KEGG), containing pathways, modules, enzymes, reactions and metabolites. In addition to providing a list of pathways, FELLA reports intermediate entities (modules, enzymes, reactions) that link the input metabolites to them. This sheds light on pathway cross talk and potential enzymes or metabolites as targets for the condition under study. FELLA has been applied to six public datasets -three from Homo sapiens, two from Danio rerio and one from Mus musculus- and has reproduced findings from the original studies and from independent literature. CONCLUSIONS: The R package FELLA offers an innovative enrichment concept starting from a list of metabolites, based on a knowledge graph representation of the KEGG database that focuses on interpretability. Besides reporting a list of pathways, FELLA suggests intermediate entities that are of interest per se. Its usefulness has been shown at several molecular levels on six public datasets, including human and animal models. The user can run the enrichment analysis through a simple interactive graphical interface or programmatically. FELLA is publicly available in Bioconductor under the GPL-3 license.