Article
Oncology
Md. Mohaiminul Islam, Noman Mohammed, Yang Wang, Pingzhao Hu
Summary: Proper analysis of high-dimensional human genomic data is necessary for understanding fundamental biological questions. However, the privacy of individuals needs to be protected when handling such sensitive data. We proposed a differential privacy-based deep learning framework to address breast cancer status, cancer type classification, and drug sensitivity prediction, achieving improved performance compared to previous work.
FRONTIERS IN ONCOLOGY
(2022)
Review
Biochemical Research Methods
Susana Vinga
Summary: The development of new molecular and cell technologies is generating a large amount of data, leading to a growth in omics databases and presenting challenges for statistical learning and computational biology in health applications. Regularized optimization methods have emerged as a promising strategy to address the high dimensionality of these data and improve accuracy in building models for biological observations.
BRIEFINGS IN BIOINFORMATICS
(2021)
Article
Mathematics
Maria C. Mariani, Francis Biney, Osei K. Tweneboah
Summary: This study investigated the prognosis of breast cancer, heart disease, and prostate cancer using 10 machine learning models, showing that no particular class of models or learning style dominated the prognosis. Nonlinear models had the best predictive performance for breast cancer data.
Article
Genetics & Heredity
Zongliang Hu, Yan Zhou, Tiejun Tong
Summary: A robust variable selection algorithm based on logistic regression was developed for meta-analyzing high-dimensional datasets, using a combination of least trimmed squared estimates and hierarchical bi-level variable selection technique, to achieve more reliable results.
FRONTIERS IN GENETICS
(2021)
Article
Biotechnology & Applied Microbiology
Marina de Leeuw, Marta R. A. Matos, Lars Keld Nielsen
Summary: Kinetic models are important for understanding metabolic systems. This paper presents a detailed kinetic model for the central carbon metabolism of Escherichia coli, which has been validated and can be used for studying cellular metabolism and metabolic pathway design.
METABOLIC ENGINEERING
(2023)
Article
Physics, Multidisciplinary
Subhash R. Lele
Summary: In scientific problems, an appropriate statistical model often involves a large number of canonical parameters, and the quantities of interest are often real-valued functions of these parameters. Bayesian inference can be used to estimate a specified function of the parameters by using the posterior distribution. Frequentist inference is usually based on the profile likelihood for the parameter of interest. However, for hierarchical models, computing the profile likelihood function is difficult due to high-dimensional integration. We propose a simple computational method using data doubling to calculate the profile likelihood for any specified function of the parameters, and provide a mathematical proof for its validity under certain conditions.
Article
Biochemical Research Methods
Yifang Wei, Lingmei Li, Xin Zhao, Haitao Yang, Jian Sa, Hongyan Cao, Yuehua Cui
Summary: Differentiating cancer subtypes is important for personalized treatment and improved prognosis. In this study, a hierarchical multi-kernel learning (hMKL) approach was proposed to identify cancer subtypes using a two-stage kernel learning strategy. The hMKL method outperformed previous methods in handling data heterogeneity and accurately estimating the number of clusters. The application of hMKL to real data sets successfully identified meaningful subtypes and key cancer-associated biomarkers. This method provides a novel toolkit for integrating heterogeneous multi-omics data and identifying cancer subtypes.
BRIEFINGS IN BIOINFORMATICS
(2023)
Article
Biochemical Research Methods
Yifang Wei, Lingmei Li, Xin Zhao, Haitao Yang, Jian Sa, Hongyan Cao, Yuehua Cui
Summary: Differentiating cancer subtypes is crucial for personalized treatment and prognosis improvement. Integrating multiomics data can provide a comprehensive understanding of cancer biological processes. The proposed hierarchical multi-kernel learning (hMKL) method outperforms traditional methods in dealing with data heterogeneity and identifying meaningful subtypes and key cancer-associated biomarkers.
BRIEFINGS IN BIOINFORMATICS
(2022)
Article
Biochemical Research Methods
Bridget A. Tripp, Hasan H. Otu
Summary: This study introduces an algorithm called OBaNK, which utilizes Bayesian networks and external knowledge to model interactions between heterogeneous high-dimensional biological data, aiming to elucidate complex functional clusters and emergent relationships associated with an observed phenotype. The results demonstrate that OBaNK successfully learns accurate interaction networks from data integrating external knowledge and identifies heterogeneous functional networks from real data.
CURRENT BIOINFORMATICS
(2022)
Review
Genetics & Heredity
Jarod Rutledge, Hamilton Oh, Tony Wyss-Coray
Summary: Age is a significant risk factor for diseases and disabilities in the elderly. Efforts to address age-related health issues require measures of biological age and the rate of aging at the molecular level. Recent advances in high-throughput omics technologies have provided tools to quantify aging at the molecular level and identify new biomarkers of biological aging through machine learning.
NATURE REVIEWS GENETICS
(2022)
Article
Computer Science, Information Systems
Maria Martinez-Garcia, Pablo Olmos
Summary: The development of high-throughput technologies has led to an increase in the dimensionality of genomics datasets, which poses a challenge for machine learning methods. In this article, the authors propose a deep latent space model for classification and dimensionality reduction, specifically addressing the issues of missing data and limited observations. The proposed model, called Deep Bayesian Logistic Regression (DBLR), produces informative low-dimensional representations, outperforms baseline methods in classification, and can handle missing entries effectively.
IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS
(2023)
Review
Biochemistry & Molecular Biology
Seong Beom Cho
Summary: Endometriosis is a common gynecological disorder in women of reproductive age, characterized by symptoms such as dysmenorrhea, irregular menstruation, and infertility. The pathogenesis of endometriosis remains unclear, but omics experiments have provided valuable insights into the molecular mechanisms involved.
Article
Immunology
Mengyuan Li, Xuejiao Gao, Xiaosheng Wang
Summary: By analyzing multi-omics datasets, we identified molecular and clinical features associated with tumor mutation burden (TMB), which could serve as valuable predictors for TMB and immunotherapy response, with potential clinical implications for cancer management.
FRONTIERS IN IMMUNOLOGY
(2023)
Article
Computer Science, Artificial Intelligence
Rami Al-Hmouz, Witold Pedrycz, Ahmed Chiheb Ammari, Ahmed Al-Hmouz
Summary: In this study, the level of difficulty in designing a model based on training data is discussed, and a variability index is proposed to quantify the nature of data. The index is model-neutral and can describe and quantify the modeling challenge irrespective of the specific model. Additionally, a method of reducing the variability index through nonlinear transformation and fuzzy rule-based model is introduced, and the concept of adversarial data is quantified using granular features.
INTERNATIONAL JOURNAL OF INTELLIGENT SYSTEMS
(2022)
Article
Biochemistry & Molecular Biology
Maria Jose Jimenez-Santos, Alba Nogueira-Rodriguez, Elena Pineiro-Yanez, Hugo Lopez-Fernandez, Santiago Garcia-Martin, Paula Gomez-Plana, Miguel Reboiro-Jato, Gonzalo Gomez-Lopez, Daniel Glez-Pena, Fatima Al-Shahrour
Summary: PanDrugs2 is an upgraded version of PanDrugs that helps researchers interpret tumor molecular alterations and guide personalized treatment selection. It incorporates multi-omics analysis and considers cancer genetic dependencies, providing therapeutic options for genes that are difficult to target. PanDrugs2 generates an intuitive report to support clinical decision-making. The database has been updated, integrating 23 primary sources that support >74K drug-gene associations obtained from 4642 genes and 14 659 unique compounds. It is freely available for use.
NUCLEIC ACIDS RESEARCH
(2023)