Journal of Computational Biophysics and Chemistry

Cell trajectory inference is very important to understand the details of tissue cell development, state differentiation and gene dynamic regulation. However, due to the high noise and heterogeneity of the single-cell data, it is challenging to infer cell trajectory in complex biological processes. Here, we proposed a new trajectory inference method, called Metric Learning Bhattacharyya Kernel Feature Decomposition (MLBKFD). In MLBKFD, a statistical model was used to infer cell trajectory by calculating the Markov transition matrix and Bhattacharyya kernel matrix between cells. Before that, to expedite the matrix calculation in the statistical model, a deep feedforward neural network was used to perform dimensionality reduction on single-cell data. The MLBKFD was evaluated on four typical datasets as well as seven recent human fetal lung datasets. Comparisons with the two outstanding methods (i.e., DTFLOW and MARGARET) demonstrate that the MLBKFD is capable of accurately inferring cell development and differentiation trajectories from single-cell data with different sizes and sources. Notably, MLBKFD exhibits nearly twice the speed of DTFLOW while maintaining high precision, particularly when dealing with large datasets. MLBKFD provides accurate and efficient trajectory inference, empowering researchers to gain deeper insights into the complex dynamics of cell development and differentiation.

Magnitude, similar to concepts like volume, cardinality or Euler characteristic, has become a key focus in combinatorics and topology. Recent advancements in topological data analysis and persistent homology have emphasized its importance. Persistent magnitude, a newly highlighted invariant introduced by Govc and Hepworth, has emerged as a notable subject of interest. In this work, we apply persistent magnitude to analyze and predict the stability of closo-carborane structures. First, we assess the stability of carboranes by employing cross-validation with different magnitude features. The Pearson correlation coefficients for stability predictions using three distinct magnitude features are 0.900, 0.882 and 0.883, respectively. These results are comparable to the Pearson correlation coefficient of 0.881 obtained when using a single feature based on persistent homology. Second, the utilization of magnitude features to predict the HOMO, LUMO and HOMO–LUMO gaps of carboranes involves conducting eight gradient boosting regressions for each scenario. The lowest correlation coefficients observed are 0.9056, 0.9385 and 0.9427, respectively. These findings highlight the promising performance of persistent magnitude features in the analysis of material structure and stability.

Raman spectroscopy is a popular technology for material identification, but it encounters difficulties when distinguishing components with closely related substances by intuition and experience. The distinction of fats and oils is a typical example, which is significant in the food industry. In this work, the Raman spectroscopy and deep learning algorithm are combined to analyze closely related animal fats (lard, butter, mutton fat and chicken fat) and vegetable oils (soybean oil and peanut oil) in a dataset. A deep neural network founded on the VGG architecture with attention mechanism is developed, reaching an accuracy of 100% for fats and oils classification. By combining Raman spectroscopy with deep learning, this research provides a potent technique for tackling the identification of similar substances.

Genetically identical cell populations growing in the same environment show a large degree of heterogeneity in gene expression profiles, resulting in significant phenotypic consequences. One of the important sources of this heterogeneity is transcriptional bursting. In recent years, single cell transcriptome sequencing (scRNA-seq) data have been widely used to infer the kinetics of transcriptional bursting. Moreover, the increasing evidence showed that genes are jointly regulated by several competitive pathways when activated by external signals, and there is molecular memory in the process of gene state switching. Therefore, a gene expression model considering both gene activation pathway and molecular memory can better reflect the nature of transcriptional bursting. In this paper, we used an approximate Bayesian computation algorithm called ABC-PRC combined with a gene expression model that considered crosstalk and memory, to infer the kinetics of transcriptional bursting. We analyze the internal relationship between transcriptional bursting and gene expression using scRNA-seq data obtained from mouse embryonic stem cells and mouse embryonic fibroblasts. Both model analysis and data simulations demonstrate that the bimodal coefficient increases with the increase of burst size, and the noise intensity decreases with the increase of burst frequency. Additionally, we show that some previously proposed models can be simplified as special cases of the gene expression model and inference algorithm proposed here under certain circumstances.

Enzymes are protein molecules that play a crucial role in various biological processes in living organisms. They function as catalysts in biological reactions such as digestion, metabolism, DNA replication and other physiological processes. Furthermore, enzymes are widely used in food production, pharmaceuticals and biofuel production. In these industries, they accelerate desired chemical reactions as biocatalysts. Therefore, applying computational methods and data-driven algorithms to predict enzyme properties is essential. Over the past decade, deep learning has made remarkable advancements in science and technology. Deep learning is a subset of machine learning algorithms that rely on artificial neural networks. These algorithms can be employed for supervised, semi-supervised and unsupervised learning. Here, to provide an update on the current literature, we provide an overview of various deep learning algorithms and recent advancements in their application to enzyme science. These applications can generally be categorized into diverse subjects: function prediction, enzyme kinetic parameters prediction, enzyme-substrate identification, condition optimization, thermophilic property prediction, enzyme catalytic site prediction and enzyme design. In conclusion, we discuss the convergence of enzyme science and deep learning, highlighting the potential opportunities and challenges.

Papillary thyroid carcinoma (PTC) is typically an indolent cancer, yet a minority of cases develop lymph node metastasis. Due to the unclear mechanisms of lymph node metastasis, a considerable number of patients undergo unnecessary surgeries. Currently, the identification of key genetic biomarkers in high-dimensional data presents a significant challenge, thereby limiting research progress in this area. Here, we proposed a hybrid filter-wrapper feature selection strategy for core factor detection and developed MethyAE, a metastasis prediction model based on DNA methylation, utilizing an end-to-end learning auto-encoder. 46 methylated CpG sites were successfully identified as crucial biomarkers for lymph node metastasis. Leveraging 447 PTC samples from the Cancer Genome Atlas (221 with metastasis, 226 without), the MethyAE model achieves 88.9% accuracy and a recall rate of 88.6% in predicting lymph node metastasis, outperforming commonly used machine learning methods like logistic regression and random forest. Furthermore, the MethyAE model exhibits favorable performance in DNA methylation data from colon cancer, bladder cancer, and breast cancer. To the best of our knowledge, this is the first attempt to predict PTC lymph node metastasis through DNA methylation, offering pivotal decision-making criteria for avoiding unnecessary surgeries and selecting appropriate treatment plans for a substantial cohort of PTC patients.

Analyzing protein–protein interaction (PPI) networks using machine learning and deep learning algorithms, alongside centrality measures, holds paramount importance in understanding complex biological systems. These advanced computational techniques enable the extraction of valuable insights from intricate network structures, shedding light on the functional relationships between proteins. By leveraging AI-driven approaches, researchers can uncover key regulatory mechanisms, identify critical nodes within the network and predict novel protein interactions with high accuracy. Ultimately, this integration of computational methodologies enhances our ability to comprehend the dynamic behavior of biological systems at a molecular level, paving the way for advancements in drug discovery, disease understanding and personalized medicine. This review paper starts by outlining various popular available PPI network databases and network centrality calculation tools. A thorough classification of various centrality measures has been identified. It primarily delves into the centrality-driven discoveries within PPI networks in biological systems and suggests using edge centrality measures and a hybrid version of node and edge centrality measures in machine learning algorithms and deep learning algorithms to predict hidden knowledge much more effectively.

The design of compounds selectively binding to specific isoforms of histone deacetylases (HDACs) is ongoing research to prevent adverse side effects. Two of the most studied isoforms are HDAC1 and HDAC6 which are important targets in various disease conditions. Here, various machine learning (ML) approaches were built and tested to predict the bioactivity and selectivity towards specific isoforms. The selectivities of compounds were precited using two different approaches: selectivity profiling and selectivity window approaches. The bioactivity models of HDAC1 and HDAC6 were used to determine the selectivities of compounds in the selectivity profiling approach. In the selectivity window approach, models were developed by directly training on the bioactivity differences of tested compounds against HDAC1 and HDAC6. In this study, firstly all available classification and regression ML algorithms in Python package were tested to find suitable algorithms. Five ML algorithms were selected based on their performances and algorithm differences. These models were compared to each other by using traditional evaluation metrics. Then, a consensus approach of the selected ML models was employed to compare the selectivity window and selectivity profiling approaches regarding their ability to distinguish HDAC1- and HDAC6-selective compounds from others. Here, the performances were also tested against an external set and it was seen that the selectivity window approach was performing slightly better in categorizing selective compounds than the selectivity profiling approach. The approaches presented in this study could be an important step to utilize for screening molecular libraries to discover selective inhibitors for targets.