RESUMEN
Deep learning is rapidly advancing many areas of science and technology with multiple success stories in image, text, voice and video recognition, robotics, and autonomous driving. In this paper we demonstrate how deep neural networks (DNN) trained on large transcriptional response data sets can classify various drugs to therapeutic categories solely based on their transcriptional profiles. We used the perturbation samples of 678 drugs across A549, MCF-7, and PC-3 cell lines from the LINCS Project and linked those to 12 therapeutic use categories derived from MeSH. To train the DNN, we utilized both gene level transcriptomic data and transcriptomic data processed using a pathway activation scoring algorithm, for a pooled data set of samples perturbed with different concentrations of the drug for 6 and 24 hours. In both pathway and gene level classification, DNN achieved high classification accuracy and convincingly outperformed the support vector machine (SVM) model on every multiclass classification problem, however, models based on pathway level data performed significantly better. For the first time we demonstrate a deep learning neural net trained on transcriptomic data to recognize pharmacological properties of multiple drugs across different biological systems and conditions. We also propose using deep neural net confusion matrices for drug repositioning. This work is a proof of principle for applying deep learning to drug discovery and development.
Asunto(s)
Algoritmos , Reposicionamiento de Medicamentos/métodos , Células A549 , Descubrimiento de Drogas , Humanos , Células MCF-7 , Redes Neurales de la Computación , Máquina de Vectores de Soporte , Transcriptoma/genéticaRESUMEN
Multi-modal data analysis techniques, such as the Parallel Independent Component Analysis (pICA), are essential in neuroscience, medical imaging and genetic studies. The pICA algorithm allows the simultaneous decomposition of up to two data modalities achieving better performance than separate ICA decompositions and enabling the discovery of links between modalities. However, advances in data acquisition techniques facilitate the collection of more than two data modalities from each subject. Examples of commonly measured modalities include genetic information, structural magnetic resonance imaging (MRI) and functional MRI. In order to take full advantage of the available data, this work extends the pICA approach to incorporate three modalities in one comprehensive analysis. Simulations demonstrate the three-way pICA performance in identifying pairwise links between modalities and estimating independent components which more closely resemble the true sources than components found by pICA or separate ICA analyses. In addition, the three-way pICA algorithm is applied to real experimental data obtained from a study that investigate genetic effects on alcohol dependence. Considered data modalities include functional MRI (contrast images during alcohol exposure paradigm), gray matter concentration images from structural MRI and genetic single nucleotide polymorphism (SNP). The three-way pICA approach identified links between a SNP component (pointing to brain function and mental disorder associated genes, including BDNF, GRIN2B and NRG1), a functional component related to increased activation in the precuneus area, and a gray matter component comprising part of the default mode network and the caudate. Although such findings need further verification, the simulation and in-vivo results validate the three-way pICA algorithm presented here as a useful tool in biomedical data fusion applications.
Asunto(s)
Algoritmos , Encéfalo/anatomía & histología , Encéfalo/fisiología , Genotipo , Modelos Estadísticos , Adulto , Alcoholismo/genética , Alcoholismo/patología , Mapeo Encefálico , Simulación por Computador , Humanos , Imagen por Resonancia Magnética , Masculino , Persona de Mediana Edad , Polimorfismo de Nucleótido Simple , Adulto JovenRESUMEN
BACKGROUND: Copy number variations (CNVs) are structural genetic mutations consisting of segmental gains or losses in DNA sequence. Although CNVs contribute substantially to genomic variation, few genetic and imaging studies report association of CNVs with alcohol dependence (AD). Our purpose is to find evidence of this association across ethnic populations and genders. This work is the first AD-CNV study across ethnic groups and the first to include the African American (AA) population. METHODS: This study considers 2 CNV data sets, one for discovery (2,345 samples) and the other for validation (239 samples), both including subjects with AD and healthy controls of European and African ancestry. Our analysis assesses the association between AD and CNV losses across ethnic groups and gender by examining the effect of overall losses across the whole genome, collective losses within individual cytogenetic bands, and specific losses in CNV regions. RESULTS: Results from the discovery data set showed an association between CNV losses within 16q12.2 and AD diagnosis (p = 4.53 × 10(-3) ). An overlapping CNV region from the validation data set exhibited the same direction of effect with respect to AD (p = 0.051). This CNV region affects the genes CES1p1 and CES1, which are members of the carboxylesterase (CES) family. The enzyme encoded by CES1 is a major liver enzyme that typically catalyzes the decomposition of ester into alcohol and carboxylic acid and is involved in drug or xenobiotics, fatty acid, and cholesterol metabolisms. In addition, the most significantly associated CNV region was located at 9p21.2 (p = 1.9 × 10(-3) ) in our discovery data set. Although not observed in the validation data set, probably due to small sample size, this result might hold potential connection to AD given its connection with neuronal death. In contrast, we did not find any association between AD and the overall total losses or the collective losses within individual cytogenetic bands. CONCLUSIONS: Overall, our study provides evidence that the specific CNVs at 16q12.2 contribute to the development of alcoholism in AA and European American populations.
Asunto(s)
Alcoholismo/complicaciones , Negro o Afroamericano/genética , Variaciones en el Número de Copia de ADN/efectos de los fármacos , Población Blanca/genética , Adulto , Alcoholismo/genética , Estudios de Casos y Controles , Cromosomas Humanos Par 16/efectos de los fármacos , Cromosomas Humanos Par 16/genética , Variaciones en el Número de Copia de ADN/genética , Etnicidad/genética , Estudio de Asociación del Genoma Completo , Genotipo , Humanos , Masculino , Estados UnidosRESUMEN
PURPOSE: The development of objective biomarkers for mild traumatic brain injury (mTBI) in the chronic period is an important clinical and research goal. Head trauma is known to affect the mechanisms that support the electrophysiological processing of information within and between brain regions, so methods like quantitative EEG may provide viable indices of brain dysfunction associated with even mTBI. METHODS: Resting-state, eyes-closed EEG data were obtained from 71 individuals with military-related mTBI and 82 normal comparison subjects without traumatic brain injury. All mTBI subjects were in the chronic period of injury (>5 months since the time of injury). Quantitative metrics included absolute and relative power in delta, theta, alpha, beta, high beta, and gamma bands, plus a measure of interhemispheric coherence in each band. Data were analyzed using univariate and multivariate methods, the latter coupled to machine learning strategies. RESULTS: Analyses revealed significant (P < 0.05) group level differences in global relative theta power (increased for mTBI patients), global relative alpha power (decreased for mTBI patients), and global beta-band interhemispheric coherence (decreased for mTBI patients). Single variables were limited in their ability to predict group membership (e.g., mTBI vs. control) for individual subjects, each with a predictive accuracy that was below 60%. In contrast, the combination of a multivariate approach with machine learning methods yielded a composite metric that provided an overall predictive accuracy of 75% for correct classification of individual subjects as coming from control versus mTBI groups. CONCLUSIONS: This study indicates that quantitative EEG methods may be useful in the identification, classification, and tracking of individual subjects with mTBI.
Asunto(s)
Conmoción Encefálica/diagnóstico , Conmoción Encefálica/fisiopatología , Electroencefalografía/métodos , Adulto , Encéfalo/fisiopatología , Femenino , Humanos , MasculinoRESUMEN
OBJECTIVES: The goal of this study was to use machine learning to more accurately predict survival after echocardiography. BACKGROUND: Predicting patient outcomes (e.g., survival) following echocardiography is primarily based on ejection fraction (EF) and comorbidities. However, there may be significant predictive information within additional echocardiography-derived measurements combined with clinical electronic health record data. METHODS: Mortality was studied in 171,510 unselected patients who underwent 331,317 echocardiograms in a large regional health system. The authors investigated the predictive performance of nonlinear machine learning models compared with that of linear logistic regression models using 3 different inputs: 1) clinical variables, including 90 cardiovascular-relevant International Classification of Diseases, Tenth Revision, codes, and age, sex, height, weight, heart rate, blood pressures, low-density lipoprotein, high-density lipoprotein, and smoking; 2) clinical variables plus physician-reported EF; and 3) clinical variables and EF, plus 57 additional echocardiographic measurements. Missing data were imputed with a multivariate imputation by using a chained equations algorithm (MICE). The authors compared models versus each other and baseline clinical scoring systems by using a mean area under the curve (AUC) over 10 cross-validation folds and across 10 survival durations (6 to 60 months). RESULTS: Machine learning models achieved significantly higher prediction accuracy (all AUC >0.82) over common clinical risk scores (AUC = 0.61 to 0.79), with the nonlinear random forest models outperforming logistic regression (p < 0.01). The random forest model including all echocardiographic measurements yielded the highest prediction accuracy (p < 0.01 across all models and survival durations). Only 10 variables were needed to achieve 96% of the maximum prediction accuracy, with 6 of these variables being derived from echocardiography. Tricuspid regurgitation velocity was more predictive of survival than LVEF. In a subset of studies with complete data for the top 10 variables, multivariate imputation by chained equations yielded slightly reduced predictive accuracies (difference in AUC of 0.003) compared with the original data. CONCLUSIONS: Machine learning can fully utilize large combinations of disparate input variables to predict survival after echocardiography with superior accuracy.
Asunto(s)
Minería de Datos/métodos , Bases de Datos Factuales , Ecocardiografía , Registros Electrónicos de Salud , Cardiopatías/diagnóstico por imagen , Aprendizaje Automático , Cardiopatías/mortalidad , Humanos , Valor Predictivo de las Pruebas , Pronóstico , Reproducibilidad de los Resultados , Estudios Retrospectivos , Medición de Riesgo , Factores de Riesgo , Factores de TiempoRESUMEN
In the biomedical field, current technology allows for the collection of multiple data modalities from the same subject. In consequence, there is an increasing interest for methods to analyze multi-modal data sets. Methods based on independent component analysis have proven to be effective in jointly analyzing multiple modalities, including brain imaging and genetic data. This paper describes a new algorithm, three-way parallel independent component analysis (3pICA), for jointly identifying genomic loci associated with brain function and structure. The proposed algorithm relies on the use of multi-objective optimization methods to identify correlations among the modalities and maximally independent sources within modality. We test the robustness of the proposed approach by varying the effect size, cross-modality correlation, noise level, and dimensionality of the data. Simulation results suggest that 3p-ICA is robust to data with SNR levels from 0 to 10 dB and effect-sizes from 0 to 3, while presenting its best performance with high cross-modality correlations, and more than one subject per 1,000 variables. In an experimental study with 112 human subjects, the method identified links between a genetic component (pointing to brain function and mental disorder associated genes, including PPP3CC, KCNQ5, and CYP7B1), a functional component related to signal decreases in the default mode network during the task, and a brain structure component indicating increases of gray matter in brain regions of the default mode region. Although such findings need further replication, the simulation and in-vivo results validate the three-way parallel ICA algorithm presented here as a useful tool in biomedical data decomposition applications.
Asunto(s)
Mapeo Encefálico/métodos , Encéfalo/fisiología , Encéfalo/fisiopatología , Imagen por Resonancia Magnética/métodos , Polimorfismo de Nucleótido Simple , Algoritmos , Simulación por Computador , Diagnóstico por Imagen , Sustancia Gris , Humanos , Procesamiento de Imagen Asistido por Computador/métodos , Distribución Normal , Análisis de Componente PrincipalRESUMEN
High data dimensionality poses a major challenge for imaging genomic studies. To address this issue, a semi-blind multivariate approach, parallel independent component analysis with multiple references (pICA-MR), is proposed. pICA-MR extracts imaging and genetic components in parallel and enhances inter-modality correlations. Prior knowledge is incorporated to emphasize genetic factors with specific attributes. Particularly, pICA-MR can investigate multiple genetic references to explore functional interactions among genes. Simulations demonstrate robust performances with Euclidean distance employed as a metric for reference similarity, where components pointed by the same references are reliably identified and the detection power is significantly improved compared to blind methods.
Asunto(s)
Estudios de Asociación Genética , Simulación por Computador , Humanos , Desequilibrio de Ligamiento , Imagen por Resonancia Magnética , Modelos Genéticos , Análisis Multivariante , Polimorfismo de Nucleótido Simple , Sensibilidad y Especificidad , Estadística como AsuntoRESUMEN
There is a growing interest in identifying neuroimaging-based biomarkers for schizophrenia. Previous studies have shown both functional and structural brain abnormalities in schizophrenia patients. One main category of these findings consists of volumetric abnormalities in brain structure in different cortical and subcortical structures in patients' brain. However there has been little work investigating changes in the brain's functional volumes. Nor has there been work studying differences in brain networks as opposed to single regions. In this study, we investigated the volumes of functional networks as potential biomarkers. Independent component analysis was used to decompose fMRI images into maximally independent spatial maps and corresponding time-courses. Volume of functional networks was computed from subject-specific back reconstructed spatial maps. The results show that different nodes of the default-mode network exhibit volumetric abnormalities in schizophrenia patients. Interestingly these networks are larger in patients compared to controls.
Asunto(s)
Encéfalo/fisiología , Imagen por Resonancia Magnética , Esquizofrenia/fisiopatología , Mapeo Encefálico , Estudios de Casos y Controles , Enfermedad Crónica , HumanosRESUMEN
Spatially-varying signal content can be effectively modeled using amplitude modulation-frequency modulation (AM-FM) representations. The AM-FM representation allow us to extract instantaneous amplitude (IA) and instantaneous frequency (IF) components that can be used to measure non-stationary content in biomedical images and videos. This paper introduces a new method for estimating the IA and the IF based on a quasi-local method (QLM). We provide an extensive comparison of AM-FM demodulation approaches based on QLM and a quasi-eigenfunction approximation method using three different filter-banks: (i) a separable, equiripple design, (ii) a Gabor filter bank, and (iii) a directional filter bank approach based on the Contourlet transform. The results document that the use of the new QLM method with an equiripple filter bank design gave the best IF magnitude estimates for a synthetic image. The new QLM method is then applied to a multi-site schizophrenia dataset (N=307). The dataset included structure magnetic resonance images from healthy controls and patients diagnosed with schizophrenia. The IF magnitude is shown to be less sensitive to variations across sites as opposed to the standard use of SMRI images that suffered from significant dependency on the scanner configurations on different collection sites. Furthermore, the regions of interest identified through the use of the IF magnitude are in agreement with previous studies.
Asunto(s)
Esquizofrenia/diagnóstico , Adulto , Algoritmos , Encéfalo/fisiopatología , Estudios de Casos y Controles , Femenino , Humanos , Aumento de la Imagen , Interpretación de Imagen Asistida por Computador , Imagen por Resonancia Magnética , Masculino , Persona de Mediana Edad , Adulto JovenRESUMEN
The association of copy number variation (CNV) with schizophrenia has been reported with evidence of increased frequency of both rare and large CNVs. Yet, little is known about the impact of CNVs in brain structure. In this pilot study, we explored collective effects of all CNVs in each cytogenetic band on the risk of schizophrenia and gray matter variation measured in structural magnetic resonance imaging. With 324 participants' CNV profiles (151 schizophrenia patients and 173 healthy controls), we first extracted specific CNV features that differ between patients and controls using a two sample t-test, and then tested their associations with gray matter concentration using a linear regression model in a subset of 301 participants. Our data first provided evidence of population structure in CNV features where elevated rare CNV burden in schizophrenia patients was confounded by the levels associated with African American subjects. We considered this ethnic group difference in the following cytoband analyses. Deletions in one cytoband 22q13.31 were observed significantly (p<0.05) more in patients than controls from all samples after controlling ethnicity, and the deletion load was also significantly (pâ=â1.44×10â»4) associated with reduced gray matter concentration of a brain network mainly comprised of the cingulate gyrus and insula. Since 80% deletion carriers were patients, patients with deletions also showed reduced gray matter concentration compared with patients without deletions (pâ=â3.36×10â»4). Our findings indicate that regional CNVs at 22q13.31, no matter the size, may influence the risk of schizophrenia with a remarkably increased mutation rate and with reduced gray matter concentration in the peri-limbic cortex. This proof-of-concept study suggests that the CNVs occurring at some 'hotspots' may in fact cause biological downstream effects and larger studies are important for confirming our initial results.
Asunto(s)
Encéfalo/patología , Deleción Cromosómica , Cromosomas Humanos Par 22 , Esquizofrenia/genética , Esquizofrenia/patología , Adulto , Estudios de Casos y Controles , Trastornos de los Cromosomas/complicaciones , Trastornos de los Cromosomas/genética , Trastornos de los Cromosomas/patología , Cromosomas Humanos Par 22/genética , Variaciones en el Número de Copia de ADN , Femenino , Frecuencia de los Genes , Humanos , Imagen por Resonancia Magnética , Masculino , Persona de Mediana Edad , Proyectos Piloto , Esquizofrenia/epidemiologíaRESUMEN
Estudio antropológico exploratorio sobre las representaciones , actitudes y prácticas relacionadas con la Leishmaniasis Cutánea(LC) en la población rural del cantón de Acosta, Costa Rica, dirigido a estimar la posibilidad de aplición de medidas de control propuestas sobre una base epidemiológica. Entrevistas abiertas con una pequena muestra de personas, provenientes tanto de casas-caso como de casas-control, proveyeron la base para un Análisis Proposicional del Discurso(APD). Los resultados son que la gente de Acosta considera la LC como una entidad nosológica diferenciada, pero se interesan sobre todo en sus manifestaciones clínicas en los ninos(que son los mas afectados), así como en su propia capacidad de acción mediante remedios populares. La idea de medidas de control sobre los reservorios, los vectores o el contexto espacio-temporal del contacto no asoma espontáneamente en el pensamiento de la gente. Sin embargo, la LC se percibe como una disrupción en el espacio doméstico y peridoméstico, considerado seguro; medidas de control que intervegan en ese ámbito podrían pues tener buenas posibilidades de éxito.