Frontiers in bioinformatics最新文献

A common practice in molecular systematics is to infer phylogeny and then scale it to time by using a relaxed clock method and calibrations. This sequential analysis practice ignores the effect of phylogenetic uncertainty on divergence time estimates and their confidence/credibility intervals. An alternative is to infer phylogeny and times jointly to incorporate phylogenetic errors into molecular dating. We compared the performance of these two alternatives in reconstructing evolutionary timetrees using computer-simulated and empirical datasets. We found sequential and joint analyses to produce similar divergence times and phylogenetic relationships, except for some nodes in particular cases. The joint inference performed better when the phylogeny was not well resolved, situations in which the joint inference should be preferred. However, joint inference can be infeasible for large datasets because available Bayesian methods are computationally burdensome. We present an alternative approach for joint inference that combines the bag of little bootstraps, maximum likelihood, and RelTime approaches for simultaneously inferring evolutionary relationships, divergence times, and confidence intervals, incorporating phylogeny uncertainty. The new method alleviates the high computational burden imposed by Bayesian methods while achieving a similar result.

Introduction: Existing large-scale preclinical cancer drug response databases provide us with a great opportunity to identify and predict potentially effective drugs to combat cancers. Deep learning models built on these databases have been developed and applied to tackle the cancer drug-response prediction task. Their prediction has been demonstrated to significantly outperform traditional machine learning methods. However, due to the "black box" characteristic, biologically faithful explanations are hardly derived from these deep learning models. Interpretable deep learning models that rely on visible neural networks (VNNs) have been proposed to provide biological justification for the predicted outcomes. However, their performance does not meet the expectation to be applied in clinical practice. Methods: In this paper, we develop an XMR model, an eXplainable Multimodal neural network for drug Response prediction. XMR is a new compact multimodal neural network consisting of two sub-networks: a visible neural network for learning genomic features and a graph neural network (GNN) for learning drugs' structural features. Both sub-networks are integrated into a multimodal fusion layer to model the drug response for the given gene mutations and the drug's molecular structures. Furthermore, a pruning approach is applied to provide better interpretations of the XMR model. We use five pathway hierarchies (cell cycle, DNA repair, diseases, signal transduction, and metabolism), which are obtained from the Reactome Pathway Database, as the architecture of VNN for our XMR model to predict drug responses of triple negative breast cancer. Results: We find that our model outperforms other state-of-the-art interpretable deep learning models in terms of predictive performance. In addition, our model can provide biological insights into explaining drug responses for triple-negative breast cancer. Discussion: Overall, combining both VNN and GNN in a multimodal fusion layer, XMR captures key genomic and molecular features and offers reasonable interpretability in biology, thereby better predicting drug responses in cancer patients. Our model would also benefit personalized cancer therapy in the future.

Introduction: The application of RNA-sequencing has led to numerous breakthroughs related to investigating gene expression levels in complex biological systems. Among these are knowledge of how organisms, such as the vertebrate model organism zebrafish (Danio rerio), respond to toxicant exposure. Recently, the development of 3' RNA-seq has allowed for the determination of gene expression levels with a fraction of the required reads compared to standard RNA-seq. While 3' RNA-seq has many advantages, a comparison to standard RNA-seq has not been performed in the context of whole organism toxicity and sparse data. Methods and results: Here, we examined samples from zebrafish exposed to perfluorobutane sulfonamide (FBSA) with either 3' or standard RNA-seq to determine the advantages of each with regards to the identification of functionally enriched pathways. We found that 3' and standard RNA-seq showed specific advantages when focusing on annotated or unannotated regions of the genome. We also found that standard RNA-seq identified more differentially expressed genes (DEGs), but that this advantage disappeared under conditions of sparse data. We also found that standard RNA-seq had a significant advantage in identifying functionally enriched pathways via analysis of DEG lists but that this advantage was minimal when identifying pathways via gene set enrichment analysis of all genes. Conclusions: These results show that each approach has experimental conditions where they may be advantageous. Our observations can help guide others in the choice of 3' RNA-seq vs standard RNA sequencing to query gene expression levels in a range of biological systems.

Traditional staining of biological specimens for microscopic imaging entails time-consuming, laborious, and costly procedures, in addition to producing inconsistent labeling and causing irreversible sample damage. In recent years, computational "virtual" staining using deep learning techniques has evolved into a robust and comprehensive application for streamlining the staining process without typical histochemical staining-related drawbacks. Such virtual staining techniques can also be combined with neural networks designed to correct various microscopy aberrations, such as out-of-focus or motion blur artifacts, and improve upon diffracted-limited resolution. Here, we highlight how such methods lead to a host of new opportunities that can significantly improve both sample preparation and imaging in biomedical microscopy.

As biological imaging continues to rapidly advance, it results in increasingly complex image data, necessitating a reevaluation of conventional bioimage analysis methods and their accessibility. This perspective underscores our belief that a transition from desktop-based tools to web-based bioimage analysis could unlock immense opportunities for improved accessibility, enhanced collaboration, and streamlined workflows. We outline the potential benefits, such as reduced local computational demands and solutions to common challenges, including software installation issues and limited reproducibility. Furthermore, we explore the present state of web-based tools, hurdles in implementation, and the significance of collective involvement from the scientific community in driving this transition. In acknowledging the potential roadblocks and complexity of data management, we suggest a combined approach of selective prototyping and large-scale workflow application for optimal usage. Embracing web-based bioimage analysis could pave the way for the life sciences community to accelerate biological research, offering a robust platform for a more collaborative, efficient, and democratized science.

Antimicrobial peptides (AMPs) are components of natural immunity against invading pathogens. They are polymers that fold into a variety of three-dimensional structures, enabling their function, with an underlying sequence that is best represented in a non-flat space. The structural data of AMPs exhibits non-Euclidean characteristics, which means that certain properties, e.g., differential manifolds, common system of coordinates, vector space structure, or translation-equivariance, along with basic operations like convolution, in non-Euclidean space are not distinctly established. Geometric deep learning (GDL) refers to a category of machine learning methods that utilize deep neural models to process and analyze data in non-Euclidean settings, such as graphs and manifolds. This emerging field seeks to expand the use of structured models to these domains. This review provides a detailed summary of the latest developments in designing and predicting AMPs utilizing GDL techniques and also discusses both current research gaps and future directions in the field.

Artificial Intelligence (AI) has achieved remarkable success in image generation, image analysis, and language modeling, making data-driven techniques increasingly relevant in practical real-world applications, promising enhanced creativity and efficiency for human users. However, the deployment of AI in high-stakes domains such as infrastructure and healthcare still raises concerns regarding algorithm accountability and safety. The emerging field of explainable AI (XAI) has made significant strides in developing interfaces that enable humans to comprehend the decisions made by data-driven models. Among these approaches, concept-based explainability stands out due to its ability to align explanations with high-level concepts familiar to users. Nonetheless, early research in adversarial machine learning has unveiled that exposing model explanations can render victim models more susceptible to attacks. This is the first study to investigate and compare the impact of concept-based explanations on the privacy of Deep Learning based AI models in the context of biomedical image analysis. An extensive privacy benchmark is conducted on three different state-of-the-art model architectures (ResNet50, NFNet, ConvNeXt) trained on two biomedical (ISIC and EyePACS) and one synthetic dataset (SCDB). The success of membership inference attacks while exposing varying degrees of attribution-based and concept-based explanations is systematically compared. The findings indicate that, in theory, concept-based explanations can potentially increase the vulnerability of a private AI system by up to 16% compared to attributions in the baseline setting. However, it is demonstrated that, in more realistic attack scenarios, the threat posed by explanations is negligible in practice. Furthermore, actionable recommendations are provided to ensure the safe deployment of concept-based XAI systems. In addition, the impact of differential privacy (DP) on the quality of concept-based explanations is explored, revealing that while negatively influencing the explanation ability, DP can have an adverse effect on the models' privacy.

Introduction: Autofluorescence imaging of the coenzymes reduced nicotinamide (phosphate) dinucleotide (NAD(P)H) and oxidized flavin adenine dinucleotide (FAD) provides a label-free method to detect cellular metabolism and phenotypes. Time-domain fluorescence lifetime data can be analyzed by exponential decay fitting to extract fluorescence lifetimes or by a fit-free phasor transformation to compute phasor coordinates. Methods: Here, fluorescence lifetime data analysis by biexponential decay curve fitting is compared with phasor coordinate analysis as input data to machine learning models to predict cell phenotypes. Glycolysis and oxidative phosphorylation of MCF7 breast cancer cells were chemically inhibited with 2-deoxy-d-glucose and sodium cyanide, respectively; and fluorescence lifetime images of NAD(P)H and FAD were obtained using a multiphoton microscope. Results: Machine learning algorithms built from either the extracted lifetime values or phasor coordinates predict MCF7 metabolism with a high accuracy (∼88%). Similarly, fluorescence lifetime images of M0, M1, and M2 macrophages were acquired and analyzed by decay fitting and phasor analysis. Machine learning models trained with features from curve fitting discriminate different macrophage phenotypes with improved performance over models trained using only phasor coordinates. Discussion: Altogether, the results demonstrate that both curve fitting and phasor analysis of autofluorescence lifetime images can be used in machine learning models for classification of cell phenotype from the lifetime data.