Article
Biochemical Research Methods
John A. Rhodes, Hector Banos, Jonathan D. Mitchell, Elizabeth S. Allman
Summary: MSCquartets is an R package for species tree hypothesis testing, inference of species trees, and inference of species networks. It takes collections of metric or topological locus trees as input, summarizes them using quartets, and displays hypothesis test results in a simplex plot. The package implements algorithms for topological and metric species tree inference, as well as level-1 topological species network inference.
Article
Computer Science, Interdisciplinary Applications
Martina Prugger, Lukas Einkemmer, Carlos F. Lopez
Summary: Solving the chemical master equation is crucial for understanding biological and chemical systems. However, directly solving it faces the curse of dimensionality. A low-rank approach based on partitioning the network into biologically relevant subsets is proposed to tackle this issue, successfully simulating large-scale biological networks.
JOURNAL OF COMPUTATIONAL PHYSICS
(2023)
Article
Computer Science, Theory & Methods
Andrew Golightly, Chris Sherlock
Summary: This study addresses the problem of inference for nonlinear, multivariate diffusion processes using incomplete, discrete-time data subject to measurement error. A novel augmentation scheme is proposed to improve inference efficiency by using correlated particle filters and an additional Gibbs step. Connections between this pseudo-marginal scheme and existing inference schemes are made to provide a unified inference framework.
STATISTICS AND COMPUTING
(2022)
Article
Computer Science, Interdisciplinary Applications
Maximilian Cohen, Dionisios G. Vlachos
Summary: The CKBIT Python software library enables users to implement Bayesian inference techniques for kinetic parameter estimation and uncertainty quantification. It leverages functionalities of other open source Python packages, offering simplified implementation through minimal user-required coding and straightforward Excel input files. The program provides various estimation options for activation energies, reaction orders, and pre-exponential terms from chemical reaction data.
COMPUTER PHYSICS COMMUNICATIONS
(2021)
Article
Nanoscience & Nanotechnology
Dexter Barrows, Silvana Ilie
Summary: This paper presents a novel method to estimate chemical reaction and diffusion rates for biochemical reaction-diffusion dynamics. The approach utilizes iterated particle filtering to fit a high-dimensional stochastic and discrete spatiotemporal model to sparse time series data, particularly in cases with low copy numbers of chemical species. The feasibility of this method is demonstrated on three realistic reaction-diffusion systems, achieving accurate recovery of known true values for all rate parameters.
Article
Management
Zice Ru, Jiapeng Liu, Milosz Kadzinski, Xiuwu Liao
Summary: This article proposes a family of probabilistic ordinal regression methods for multiple criteria sorting. It introduces Bayesian Ordinal Regression and Subjective Stochastic Ordinal Regression to derive the class assignments of alternatives based on provided preference information. The introduced approaches are evaluated through an experimental study involving real-world datasets.
EUROPEAN JOURNAL OF OPERATIONAL RESEARCH
(2023)
Article
Environmental Sciences
Peter Reichert, Lorenz Ammann, Fabrizio Fenicia
Summary: Stochastic hydrological process models have two conceptual advantages over deterministic models: providing a more realistic description of the system and better accounting for structural deficits. However, stochastic models are more susceptible to identifiability problems and Bayesian inference is computationally more demanding.
WATER RESOURCES RESEARCH
(2021)
Article
Mathematics, Applied
Oscar Bates, Lluis Guasch, George Strong, Thomas Caradoc Robins, Oscar Calderon-Agudo, Carlos Cueto, Javier Cudeiro, Mengxing Tang
Summary: Bayesian methods are popular for inverse problems research. Stochastic Variational Inference (SVI) uses gradient-based methods to solve Bayes' equation and is suitable for time-limited or computationally expensive applications. SVI can find a cost-free estimate of the pixel-wise variance of the sound-speed distribution, which can be used for uncertainty estimation in pixel-wise sound-speed reconstruction.
Article
Multidisciplinary Sciences
Kaan oecal, Michael U. Gutmann, Guido Sanguinetti, Ramon Grima
Summary: Estimating uncertainty in model predictions is a central task in quantitative biology. We propose a flexible procedure for uncertainty quantification in a wide class of reaction networks describing stochastic gene expression including those with feedback. The method is based on creating a tractable coarse-graining of the model that is learned from simulations, a synthetic model, to approximate the likelihood function. Synthetic models can substantially outperform state-of-the-art approaches, providing an accurate and computationally viable solution to uncertainty quantification in stochastic models of gene expression.
JOURNAL OF THE ROYAL SOCIETY INTERFACE
(2022)
Article
Engineering, Electrical & Electronic
Abhishek Grover, Brejesh Lall
Summary: The research focuses on modeling the chemical sensing process, analyzing binding and unbinding reactions on the sensor surface, and studying fluctuations in sensor response. By formulating the binding and unbinding reactions as Bernoulli trials, a stochastic differential equation is derived and used to analyze the frequency spectrum of the sensor signal, demonstrating the potential value of the model in agent identification applications.
IEEE TRANSACTIONS ON SIGNAL PROCESSING
(2021)
Article
Multidisciplinary Sciences
Augustinas Sukys, Kaan Ocal, Ramon Grima
Summary: The Chemical Master Equation provides an accurate description of stochastic biochemical reaction networks, but it is analytically intractable for most practical systems. This article proposes a neural network-based approach, called Nessie, to approximate the solutions of the CME using a small number of stochastic simulations. The method demonstrates the ability of simple neural networks to capture complex distributions in parameter space, enabling faster computationally intensive tasks.
Article
Automation & Control Systems
Haishan Ye, Luo Luo, Zhihua Zhang
Summary: The paper proposes a unified framework to analyze the local and global convergence properties of second order methods, bridging the gap between current convergence theory and empirical performance in real applications.
JOURNAL OF MACHINE LEARNING RESEARCH
(2021)
Article
Chemistry, Physical
Kaan oecal, Guido Sanguinetti, Ramon Grima
Summary: Model reduction is a crucial tool for quantitative biologists due to the complexity of mathematical models in biology. In this paper, we demonstrate that common approaches to model reduction for Chemical Master Equation can be viewed as minimizing the Kullback-Leibler divergence between the full model and its reduction. This enables us to transform the task of model reduction into a variational problem that can be solved using numerical optimization methods, and we also derive general expressions for propensities of a reduced system.
JOURNAL OF CHEMICAL PHYSICS
(2023)
Article
Automation & Control Systems
William J. Wilkinson, Simo Sarkka, Arno Solin
Summary: In this study, we generalize the natural gradient variational inference, expectation propagation, and posterior linearisation as Newton's method for optimizing the parameters of a Bayesian posterior distribution. Our approach provides new insights into the connections between various inference schemes and guarantees positive semi-definite covariance matrices.
JOURNAL OF MACHINE LEARNING RESEARCH
(2023)
Article
Computer Science, Artificial Intelligence
Peng Lin, Changsheng Dou, Nannan Gu, Zhiyuan Shi, Lili Ma
Summary: This paper proposes a new amortized generalized belief propagation (GBP) algorithm applied to Bayesian networks (BN) for efficient inference. The algorithm utilizes pairwise conversion and loop structured region graph algorithms to generate a valid region graph and directly amortizes the energy function using (deep) neural networks. Empirical studies show that the proposed algorithm significantly improves convergence and efficiency compared to conventional algorithms.
INTERNATIONAL JOURNAL OF INTELLIGENT SYSTEMS
(2023)
Article
Physics, Multidisciplinary
Kaan Ocal, Ramon Grima, Guido Sanguinetti
JOURNAL OF PHYSICS A-MATHEMATICAL AND THEORETICAL
(2020)
Article
Genetics & Heredity
Giulio Caravagna, Timon Heide, Marc J. Williams, Luis Zapata, Daniel Nichol, Ketevan Chkhaidze, William Cross, George D. Cresswell, Benjamin Werner, Ahmet Acar, Louis Chesler, Chris P. Barnes, Guido Sanguinetti, Trevor A. Graham, Andrea Sottoriva
Article
Genetics & Heredity
Kashyap Chhatbar, Justyna Cholewa-Waclaw, Ruth Shah, Adrian Bird, Guido Sanguinetti
Article
Biochemical Research Methods
Giulio Caravagna, Guido Sanguinetti, Trevor A. Graham, Andrea Sottoriva
BMC BIOINFORMATICS
(2020)
Review
Biochemistry & Molecular Biology
Viplove Arora, Guido Sanguinetti
Summary: RNA-protein interactions play a crucial role in gene expression regulation. The development of scalable experimental techniques has revolutionized the field and generated large-scale datasets for machine learning. This note discusses the challenges of applying machine learning in computational RNA biology, with a focus on predicting RNA-protein interactions from next-generation sequencing data.
STATISTICAL APPLICATIONS IN GENETICS AND MOLECULAR BIOLOGY
(2022)
Article
Biochemical Research Methods
Christos Maniatis, Catalina Vallejos, Guido Sanguinetti
Summary: Single-cell multi-omics assays provide unprecedented opportunities to explore epigenetic regulation at the cellular level. However, high levels of technical noise and data sparsity often result in a lack of statistical power in correlative analyses. SCRaPL is a novel computational tool that addresses this issue by carefully modeling noise in the experimental systems.
PLOS COMPUTATIONAL BIOLOGY
(2022)
Article
Chemistry, Physical
Kaan oecal, Guido Sanguinetti, Ramon Grima
Summary: Model reduction is a crucial tool for quantitative biologists due to the complexity of mathematical models in biology. In this paper, we demonstrate that common approaches to model reduction for Chemical Master Equation can be viewed as minimizing the Kullback-Leibler divergence between the full model and its reduction. This enables us to transform the task of model reduction into a variational problem that can be solved using numerical optimization methods, and we also derive general expressions for propensities of a reduced system.
JOURNAL OF CHEMICAL PHYSICS
(2023)
Proceedings Paper
Computer Science, Artificial Intelligence
Ginevra Carbone, Luca Bortolussi, Guido Sanguinetti
Summary: This article explores the stability of saliency-based explanations of Neural Network predictions under adversarial attacks. Empirical evidence shows that interpretations provided by Bayesian Neural Networks are more stable under adversarial perturbations and attacks. The article also provides a theoretical explanation based on the geometry of the data manifold.
2022 INTERNATIONAL JOINT CONFERENCE ON NEURAL NETWORKS (IJCNN)
(2022)
Article
Multidisciplinary Sciences
Kaan oecal, Michael U. Gutmann, Guido Sanguinetti, Ramon Grima
Summary: Estimating uncertainty in model predictions is a central task in quantitative biology. We propose a flexible procedure for uncertainty quantification in a wide class of reaction networks describing stochastic gene expression including those with feedback. The method is based on creating a tractable coarse-graining of the model that is learned from simulations, a synthetic model, to approximate the likelihood function. Synthetic models can substantially outperform state-of-the-art approaches, providing an accurate and computationally viable solution to uncertainty quantification in stochastic models of gene expression.
JOURNAL OF THE ROYAL SOCIETY INTERFACE
(2022)
Article
Biochemistry & Molecular Biology
Yuanhua Huang, Guido Sanguinetti
Summary: This review discusses model-based approaches in single-cell RNA sequencing analysis within the framework of Bayesian statistics, highlighting the advantages and remaining challenges in this expanding research area.
CURRENT OPINION IN SYSTEMS BIOLOGY
(2021)
Proceedings Paper
Computer Science, Artificial Intelligence
Ginevra Carbone, Guido Sanguinetti, Luca Bortolussi
Summary: In this study, two training techniques, namely RP-Ensemble and RP-Regularizer, are proposed to enhance the robustness of Neural Networks against adversarial attacks. Both methods leverage random projections to exploit dimensionality reduction and geometric properties of adversarial perturbations, aiming to improve the network's resistance to input manipulations.
2021 INTERNATIONAL JOINT CONFERENCE ON NEURAL NETWORKS (IJCNN)
(2021)
Article
Biotechnology & Applied Microbiology
Yuanhua Huang, Guido Sanguinetti
Summary: RNA splicing plays a crucial role in driving heterogeneity in single cells through alternative transcript expression and transcriptional kinetics. BRIE2, a scalable computational method, effectively identifies differential disease-associated alternative splicing events and improves RNA velocity analysis, enabling exploration of the association between splicing phenotypes and biological changes.
Article
Biotechnology & Applied Microbiology
Paolo Marangio, Ka Ying Toby Law, Guido Sanguinetti, Sander Granneman
Summary: The diffBUM-HMM model accurately detects RNA flexibility and conformational changes from high-throughput RNA structure-probing data, accounting for noise and variability. It demonstrates higher sensitivity than existing methods and is robust against false positives, showing its value in quantitatively detecting RNA structural changes and RNA-binding protein binding sites.
Article
Biotechnology & Applied Microbiology
Chantriolnt-Andreas Kapourani, Ricard Argelaguet, Guido Sanguinetti, Catalina A. Vallejos
Summary: High-throughput single-cell measurements of DNA methylomes reveal the role of methylation heterogeneity in gene regulation. The hierarchical Bayesian model scMET addresses technical limitations, quantifies biological heterogeneity, identifies highly variable epigenetic features, and facilitates the characterization of epigenetically distinct cell populations.
Proceedings Paper
Biochemical Research Methods
Ankit Gupta, Mustafa Khammash, Guido Sanguinetti
COMPUTATIONAL METHODS IN SYSTEMS BIOLOGY (CMSB 2019)
(2019)