Long Song

Treatment-induced ovarian function loss is a significant concern for many young patients with breast cancer. Accurately predicting this risk is crucial for counselling young patients and informing their fertility-related decision-making. However, current risk prediction models for treatment-related ovarian function loss have limitations. To provide a broader representation of patient cohorts and improve feature selection, we combined retrospective data from six datasets within the FoRECAsT (Infertility after Cancer Predictor) databank, including 2679 pre-menopausal women diagnosed with breast cancer. This combined dataset presented notable missingness, prompting us to employ cross imputation using the k-nearest neighbours (KNN) machine learning (ML) algorithm. Employing Lasso regression, we developed an ML model to forecast the risk of treatment-related amenorrhea as a surrogate marker of ovarian function loss at 12 months after starting chemotherapy. Our model identified 20 variables significantly associated with risk of developing amenorrhea. Internal validation resulted in an area under the receiver operating characteristic curve (AUC) of 0.820 (95% CI: 0.817–0.823), while external validation with another dataset demonstrated an AUC of 0.743 (95% CI: 0.666–0.818). A cutoff of 0.20 was chosen to achieve higher sensitivity in validation, as false negatives—patients incorrectly classified as likely to regain menses—could miss timely opportunities for fertility preservation if desired. At this threshold, internal validation yielded sensitivity and precision rates of 91.3% and 61.7%, respectively, while external validation showed 92.9% and 60.0%. Leveraging ML methodologies, we not only devised a model for personalised risk prediction of amenorrhea, demonstrating substantial enhancements over existing models but also showcased a robust framework for maximally harnessing available data sources.

Treatment-induced ovarian function loss is a significant concern for many young patients with breast cancer. Accurately predicting this risk is crucial for counselling young patients and informing their fertility-related decision-making. However, current risk-prediction models for treatment-related ovarian function loss have limitations. To provide a broader representation of patient cohorts and improve feature selection, we combined retrospective data from six datasets within the FoRECAsT (Infertility after Cancer Predictor) databank, including 2,679 pre-menopausal women diagnosed with breast cancer. This combined dataset presented notable missingness, prompting us to employ cross imputation using the k-nearest neighbours (KNN) machine learning (ML) algorithm. Employing Lasso regression, we developed an ML model to forecast the risk of treatment-related amenorrhea as a surrogate marker of ovarian function loss at 12 months after starting chemotherapy. Our model identified 20 variables significantly associated with risk of developing amenorrhea. Internal validation resulted in an area under the receiver operating characteristic curve (AUC) of 0.820 (95% CI: 0.817-0.823), while external validation with another dataset demonstrated an AUC of 0.743 (95% CI: 0.666-0.818). A cutoff of 0.20 was chosen to achieve higher sensitivity in validation, as false negatives—patients incorrectly classified as likely to regain menses—could miss timely opportunities for fertility preservation if desired. At this threshold, internal validation yielded sensitivity and precision rates of 91.3% and 61.7%, respectively, while external validation showed 92.9% and 60.0%. Leveraging ML methodologies, we not only devised a model for personalised risk prediction of amenorrhea, demonstrating substantial enhancements over existing models but also showcased a robust framework for maximally harnessing available data sources.

Objective: Machine learning (ML) models have emerged as tools to predict length of stay (LOS) and disposition decision (DD) in emergency departments (EDs) to combat overcrowding. However, site-specific ML models are not transferable to different sites. Our objective was to develop interpretable ML models to predict LOS and DD at specific time points, all while establishing a transparent data analysis framework. This framework was designed to be easily adapted by other institutions for the development of their own ML models.Methods: We analysed data from 297 392 ED visits of patients aged 18 and above at a quaternary hospital between 30 June 2019 and 31 December 2022. Eight ML algorithms were evaluated, and ultimately, twelve lasso models built from 21 features were trained to predict four outcomes of LOS and DD at three time points post-triage. Hold-out testing and cross-validation were conducted for these models.Results: The area under the curve values were 0.862/0.868/0.878 for binary LOS predictions at 10, 60 and 120-minute time points and 0.839/0.851/0.863 for binary DD predictions. The accuracies were 60.2%/60.7%/61.9% for ternary LOS predictions and 61.5%/62.3%/63.4% for ternary DD predictions.Conclusions: Interpretable ML models demonstrated outstanding performances in predicting both LOS and DD. The transparent data analysis framework can be easily adapted by other institutions.

The leaf rust pathogen, Puccinia triticina (Pt), threatens global wheat production. The deployment of leaf rust (Lr) resistance (R) genes in wheat varieties is often followed by the development of matching virulence in Pt due to presumed changes in avirulence (Avr) genes in Pt. Identifying such Avr genes is a crucial step to understand the mechanisms of wheat-rust interactions. This study is the first to develop and apply an integrated framework of gene expression, single nucleotide polymorphism (SNP), insertion/deletion (InDel), and copy number variation (CNV) analysis in a rust fungus and identify candidate avirulence genes. Using a long-read based de novo genome assembly of an isolate of Pt (‘Pt104’) as the reference, whole-genome resequencing data of 12 Pt pathotypes derived from three lineages Pt104, Pt53, and Pt76 were analyzed. Candidate avirulence genes were identified by correlating virulence profiles with small variants (SNP and InDel) and CNV, and RNA-seq data of an additional three Pt isolates to validate expression of genes encoding secreted proteins (SPs). Out of the annotated 29,043 genes, 2392 genes were selected as SP genes with detectable expression levels. Small variant comparisons between the isolates identified 27–40 candidates and CNV analysis identified 14–31 candidates for each Avr gene, which when combined, yielded the final 40, 64, and 69 candidates for AvrLr1, AvrLr15, and AvrLr24, respectively. Taken together, our results will facilitate future work on experimental validation and cloning of Avr genes. In addition, the integrated framework of data analysis that we have developed and reported provides a more comprehensive approach for Avr gene mining than is currently available.

Leaf rust, caused by Puccinia triticina (Pt), is one of the most devastating diseases of wheat, affecting production in nearly all wheat-growing regions worldwide. Despite its economic importance, genomic resources for Pt are very limited. In the present study, we have used long-read sequencing (LRS) and the pipeline of FALCON and FALCON-Unzip (v4.1.0) to carry out the first LRS-based de novo genome assembly for Pt. Using 22.4-Gb data with an average read length of 11.6 kb and average coverage of 150-fold, we generated a genome assembly for Pt104 [strain 104-2,3,(6),(7),11; isolate S423], considered to be the founding isolate of a clonal lineage of Pt in Australia. The Pt104 genome contains 162 contigs with a total length of 140.5 Mb and N₅₀ of 2 Mb, with the associated haplotigs providing haplotype information for 91% of the genome. This represents the best quality of Pt genome assembly to date, which reduces the contig number by 91-fold and improves the N₅₀ by 4-fold as compared to the previous Pt race1 assembly. An annotation pipeline that combined multiple lines of evidence including the transcriptome assemblies derived from RNA-Seq, previously identified expressed sequence tags and Pt race 1 protein sequences predicted 29,043 genes for Pt104 genome. Based on the presence of a signal peptide, no transmembrane segment, and no target location to mitochondria, 2,178 genes were identified as secreted proteins (SPs). Whole-genome sequencing (Illumina paired-end) was performed for Pt104 and six additional strains with differential virulence profile on the wheat leaf rust resistance genes Lr26, Lr2a, and Lr3ka. To identify candidates for the corresponding avirulence genes AvrLr26, AvrLr2a, and AvrLr3ka, genetic variation within each strain was first identified by mapping to the Pt104 genome. Variants within predicted SP genes between the strains were then correlated to the virulence profiles, identifying 38, 31, and 37 candidates for AvrLr26, AvrLr2a, and AvrLr3ka, respectively. The identification of these candidate genes lays a good foundation for future studies on isolating these avirulence genes, investigating the molecular mechanisms underlying host–pathogen interactions, and the development of new diagnostic tools for pathogen monitoring.