Predicting wildfires using Machine Learning models is relevant and essential to minimize wildfire threats to protect human lives and reduce significant property damage. Reliance on traditional wildfire indices for forecasting wildfires has failed to provide the expected prediction outcomes, resulting in limited application of these models. Thus, this research compares the outcome of wildfire forecasting using fire danger rating indices against Machine Learning model outcomes. Furthermore, the performance effectiveness of the fire danger rating indices and Machine Learning model outcomes are assessed using the same wildfire incidents. The One-class Machine Learning algorithms used are Support Vector Machine, Isolation Forest, Neural network-based Autoencoder and Variational Autoencoder models. The two global wildfire indices investigated were the US National Fire Danger Rating System for California and the McArthur Forest Fire Danger Index for Western Australia, using similar features. For the same data sets, the National Fire Danger Rating System and the McArthur Forest Fire Danger Index prediction outcomes were compared with Machine Learning model outcomes. Higher wildfire prediction accuracy was achieved by the One-class models, exceeding the performance of the two wildfire danger indices by at least 20%. The implications of our research findings have the potential to influence both these wildfire indices and state-of-the-art methods in wildfire prediction by proposing alternative ML methods to model the onset of wildfires.

This research reviews recent advances in deep learning approaches tailored for regulatory genomics. It highlights how computational methods help decipher complex regulatory mechanisms within non-coding genomic regions across various tissues, emphasizing predictive applications such as transcription factor binding, chromatin accessibility, RNA processes, and RNA-protein interactions. The paper also discusses the evolution from traditional neural networks to advanced models like transformers and graph neural networks, considering three-dimensional genomic structures. Despite the promising performance, it acknowledges ongoing challenges like overfitting, biological variability, and limited dataset diversity. It emphasizes the urgent need for continued development of interpretable deep learning models to improve functional genomic annotation, underlining this task’s significance in the genomics field.

Summary: Bioinformatics is a rapidly evolving field with numerous specialized tools developed for essential genomic analysis tasks, such as read simulation, mapping, and variant calling. However, managing these tools presents significant challenges due to varied dependencies, execution steps, and output formats, complicating the installation and configuration processes. To address these issues, we introduce “Mars” a bioinformatics solution encapsulated within a singularity container that preloads a comprehensive suite of widely used genomic tools. Mars not only simplifies the installation of these tools but also automates critical workflow functions, including sequence sample preparation, read simulation, read mapping, variant calling, and result comparison. By streamlining the execution of these workflows, Mars enables users to easily manage input-output formats and compare results across different tools, thereby enhancing reproducibility and efficiency. Furthermore, by providing a cohesive environment that integrates tool management with a flexible workflow interface, Mars empowers researchers to focus on their analyses rather than the complexities of tool configuration. This integrated solution facilitates the testing of various combinations of tools and algorithms, enabling users to evaluate performance based on different metrics and identify the optimal tools for their specific genomic analysis needs. Through Mars, we aim to enhance the accessibility and usability of bioinformatics tools, ultimately advancing research in genomic analysis. Availability and implementation: Mars is freely available at https://github.com/GenomicAI/mars. It is implemented within a Singularity container environment and supports modular extension for additional genomic tools and custom workflows.

Machine learning and artificial intelligence models have become popular for climate change prediction. Forested regions in California and Western Australia are increasingly facing intense wildfires, while other parts of the world face various climate-related challenges. To address these issues, machine learning and artificial intelligence models have been developed to predict wildfire risks and support mitigation strategies. Our study focuses on developing wildfire prediction models using one-class classification algorithms. These include Support Vector Machine, Isolation Forest, AutoEncoder, Variational AutoEncoder, Deep Support Vector Data Description, and Adversarially Learned Anomaly Detection. The models were validated through five-fold cross-validation to minimize bias in selecting training and testing data. The results showed that these one-class machine learning models outperformed two-class machine learning models based on the same ground truth data, achieving mean accuracy levels between 90% and 99%. Additionally, we employed Shapley values to identify the most significant features affecting the wildfire prediction models, contributing a novel perspective to wildfire prediction research. When analyzing models trained on the California dataset, seasonal maximum and mean dew point temperatures were critical factors. These insights can significantly improve wildfire mitigation strategies. Furthermore, we have made these models accessible and user-friendly by operationalizing them through a REST API using Python Flask 1.1.2 and developing a web-based tool.

This research explores the development of wildfire prediction models using One-class classification algorithms, addressing limitations in previous research that relied on biased or imbalanced datasets for multi-class classification. A range of diverse One-class algorithms were used to build linear, ensemble, proximity-based, probabilistic, and neural network-based models, using as input features from California and Western Australian fire data sets categorized by topography, weather, and plant fuel moisture. These One-class models achieved mean test accuracy rates between 55% and 99% , with deep learning-based One-class models obtaining the highest mean test accuracy. Significant contributions to our work include a comprehensive investigation of the performance of different one-class classification methods for wildfire prediction and identifying key features influencing wildfire ignition. These features include precipitation, elevation, vapour pressure deficit, and dew point temperature, computed using Shapley values. Our work demonstrates the utility of One-class modelling in advancing wildfire prediction.