Fathima Nuzla Ismail

Hantavirus pulmonary syndrome (HPS) is a severe disease first recognised in the United States in 1993, with a case fatality rate approaching 35–50%. Since the identification of Sin Nombre virus during the Four Corners outbreak, understanding the transmission dynamics and geographic distribution of hantaviruses has become critical for public health planning and prevention. This review synthesises evidence from ecological niche modelling, epidemiological surveillance, and environmental analyses to explain patterns of hantavirus occurrence in North America. The findings indicate that HPS risk is associated with dry climates, rural and peri-urban landscapes, rodent host ecology, and increased social vulnerability, with cases primarily concentrated in the western United States. We examine how environmental conditions, rodent ecology, human–rodent interactions, and socioeconomic factors interact to influence disease risk. This synthesis provides recommendations for reducing exposure in high-risk populations and regions.

The emergence of pangenome graphs represents a paradigm shift in genomics, moving beyond linear reference genomes to embrace the full spectrum of genetic diversity within and across species. These graph-based models provide a unified framework for representing alternative haplotypes, structural variants, and complex genomic rearrangements that are often missed by traditional approaches. This paper reviews the latest developments in pangenome graph construction, data structures, alignment algorithms, and variant inference. We explore recent human, plant, and microbial genomics applications, highlighting the advantages of graph representations in capturing population diversity and improving read mapping accuracy. We briefly discuss emerging directions such as machine learning-assisted graph analysis, although current applications remain limited and exploratory. Furthermore, we examine the emerging field of clinical genomics, where pangenome references have demonstrated measurable improvements in diagnostic yield—specifically increasing variant calling sensitivity for complex structural variants by up to 10–40% compared to linear models. We conclude by addressing ongoing challenges in graph scalability and standardization, aiming to guide future research and implementation in this rapidly evolving field.

Rice cultivation accounts for roughly 10% of worldwide anthropogenic greenhouse gas emissions, making it a significant source of methane (CH₄) Despite modest observational constraints, estimates of worldwide CH₄ emissions from rice agriculture range from 18–115 Tg CH₄ yr⁻¹. CH₄ is a potent greenhouse gas, and its oxidation produces tropospheric ozone (O₃), which is harmful to public health and crop production when combined with nitrogen oxides (NOx) and sunlight. Elevated O₃ levels reduce air quality, crop productivity, and human respiratory health. This study presents an AI-driven framework that combines ensemble learning, hyperparameter optimisation (HPs), and SHAP-based explainability to enhance CH₄ emission predictions from rice paddies in India, Bangladesh, and Vietnam. The model consists of two stages: (1) a classification stage to distinguish between zero and non-zero CH₄ emissions, and (2) a regression stage to estimate emission magnitudes for non-zero situations. The framework also incorporates O₃ and asthma incidence data to assess the downstream impacts of CH₄-driven ozone formation on air quality and health outcomes. Understanding the factors that drive optimal model performance and the relative importance of features affecting model outputs is still an ongoing field of research. To address these issues, we present an integrated approach that utilises recent improvements in model optimisation and employs SHapley Additive ExPlanations (SHAP) to find the most relevant variables affecting methane (CH₄) emission forecasts. In addition, we developed a web-based artificial intelligence platform to help policymakers and stakeholders with climate strategy and sustainable agriculture by visualising methane fluxes from 2018 to 2020, ensuring practical applicability. Our findings show that ensemble learning considerably improves the accuracy of CH₄ emission prediction, minimises uncertainty, and shows the wider benefits of methane reduction for climate stability, air quality, and public health.

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.

A significant advancement in bioinformatics is using genome graph techniques to improve variation discovery across organisms. Traditional approaches, such as bwa mem, rely on linear reference genomes for genomic analyses but may introduce biases when applied to highly diverse bacterial genomes of the same species. Pangenome graphs provide an alternative paradigm for evaluating structural and minor variations within a graphical framework, including insertions, deletions, and single nucleotide polymorphisms. Pangenome graphs enhance the detection and interpretation of complex genetic variants by representing the full genetic diversity of a species. In this study, we present a robust and reliable bioinformatics pipeline utilising the PanGenome Graph Builder (PGGB) and the Variation Graph toolbox (vg giraffe) to align whole-genome sequencing data, call variants against a graph reference, and construct pangenomes from assembled genomes. Our results demonstrate that leveraging pangenome graphs over a single linear reference genome significantly improves mapping rates and variant calling accuracy for simulated and actual bacterial pathogens datasets.

The exponential growth of big data has transformed how large organisations leverage information to drive innovation, optimise processes, and maintain competitive advantages. However, managing and extracting insights from vast, heterogeneous data sources requires a scalable, secure, and well-integrated big data architecture. This paper proposes a comprehensive big data framework that aligns with organisational objectives while ensuring flexibility, scalability, and governance. The architecture encompasses multiple layers, including data ingestion, transformation, storage, analytics, machine learning, and security, incorporating emerging technologies such as Generative AI (GenAI) and low-code machine learning. Cloud-based implementations across Google Cloud, AWS, and Microsoft Azure are analysed, highlighting their tools and capabilities. Additionally, this study explores advancements in big data architecture, including AI-driven automation, data mesh, and Data Ocean paradigms. By establishing a structured, adaptable framework, this research provides a foundational blueprint for large organisations to harness big data as a strategic asset effectively.

Next-generation sequencing (NGS) is a pivotal technique in genome sequencing due to its high throughput, rapid results, cost-effectiveness, and enhanced accuracy. Its significance extends across various domains, playing a crucial role in identifying genetic variations and exploring genomic complexity. NGS finds applications in diverse fields such as clinical genomics, comparative genomics, functional genomics, and metagenomics, contributing substantially to advancements in research, medicine, and scientific disciplines. Within the sphere of genomics data science, the execution of read simulation, mapping, and variant calling holds paramount importance for obtaining precise and dependable results. Given the plethora of tools available for these purposes, each employing distinct methodologies and options, a nuanced understanding of their intricacies becomes imperative for optimization. This research, situated at the intersection of data science and genomics, involves a meticulous assessment of various tools, elucidating their individual strengths and weaknesses through rigorous experimentation and analysis. This comprehensive evaluation has enabled the researchers to pinpoint the most accurate tools, reinforcing the alignment between the established workflow and the demonstrated efficacy of specific tools in the context of genomics data analysis. To meet these requirements, "VarFind", an open-source and freely accessible pipeline tool designed to automate the entire process has been introduced (VarFind GitHub repository: https://github.com/shanikawm/varfinder)

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.