Biology subjects

Mills, D.

Publications and source records attributed to Mills, D..

4 recordsLinked to original sources

bioRxiv2025

Evaluating a structured expert elicitation approach for adaptive conservation: Lessons from five years in practice

Threatened species management relies on Ex ante estimates of species responses to different interventions to generate meaningful predictions. Structured expert elicitation is often used to generate these estimates, but comparisons of these expert-predicted outcomes with observed results are rare. This study aims to evaluate the utility of expert elicitation for adaptive management in the New South Wales Saving our Species (SoS) program in Australia by revisiting six species management plans that were generated from bespoke structured elicitation guidelines five years prior. Each species management plan included a defined scope, conceptual model, monitoring indicators and estimated response to management curves under different scenarios. Experts reviewed the conceptual models after five years of management and monitoring and compared the predicted response to management with observed monitoring data. In three of the six case studies, observed outcomes closely matched predictions. Where predictions diverged, factors such as unanticipated new threats and unexpected responses to interventions contributed to discrepancies. However, in all cases, the structured approach provided a clear logic for planning, enabling managers to systematically refine their understanding. The conceptual models and response curves proved valuable for collaboration, communication, and generating hypotheses for unexpected results. This work demonstrates the value of the bespoke guidelines in supporting adaptive management processes, strengthening the knowledge base for threatened species conservation while improving alignment between predictions and real-world outcomes.

ecology↗

bioRxiv2022

The application of neural networks to classify dolphin echolocation clicks

Passive acoustic monitoring (PAM) is a common approach to monitor marine mammal populations, for species of dolphins, porpoises and whales that use sound for navigation, feeding and communication. PAM produces large datasets which benefit from the application of machine learning algorithms to automatically detect and classify the vocalisations of these animals. We present a deep learning approach for the classification of dolphins echolocation clicks into two species groups in an environment with high background noise. We compare the use of Convolutional Neural Networks (CNN) and Recurrent Neural Network (RNN), in which we feed the models the raw waveform data and spectrograms. We show that both models perform well, with the highest performance achieved by a CNN fed with spectrograms (F1 score 97 %) and an RNN fed with raw data (F1 score 96%) fitted with Gated Recurrent Units (GRU). We recommend the use of such models to classify echolocation clicks in marine environments where background noise levels exhibit high spatial and temporal variance. In particular, the RNN showed excellent performance, while being fed with raw data, in terms of reduced processing time and storage. Deep learning automatically extracts effective features from the raw waveform in the training process through multiple layers of the model, without the need to rely on feature extraction in a separate pre-processing step.

ecology↗

bioRxiv2020

The plate-to-rod transition in trabecular bone loss is elusive

Changes in trabecular micro-architecture are key to our understanding of osteoporosis. Previous work focusing on structure model index (SMI) measurements have concluded that disease progression entails a shift from plates to rods in trabecular bone, but SMI is heavily biased by bone volume fraction. As an alternative to SMI, we proposed the Ellipsoid Factor (EF) as a continuous measure of local trabecular shape between plate-like and rod-like extremes. We investigated the relationship between EF distributions, SMI and bone volume fraction of the trabecular geometry in a murine model of disuse osteoporosis as well as from human vertebrae of differing bone volume fraction. We observed a moderate shift in EF median (at later disease stages in mouse tibia) and EF mode (in the vertebral samples with low bone volume fraction) towards a more rod-like geometry, but not in EF maximum and minimum. These results support the notion that the plate to rod transition does not coincide with the onset of bone loss and is considerably more moderate, when it does occur, than SMI suggests. A variety of local shapes not straightforward to categorise as rod or plate exist in all our trabecular bone samples.

physiology↗

bioRxiv2019

Cervicovaginal microbiome composition drives metabolic profiles in healthy pregnancy

BackgroundMicrobes and their metabolic products influence early-life immune and microbiome development, yet remain understudied during pregnancy. Vaginal microbial communities are typically dominated by one or a few well adapted microbes, which are able to survive in a narrow pH range. In comparison to other human-associated microbes, vaginal microbes are adapted to live on host-derived carbon sources, likely sourced from glycogen and mucin present in the vaginal environment. MethodsUsing 16S rRNA and ITS amplicon sequencing, we characterized the cervicovaginal microbiomes of 18 healthy women throughout the three trimesters of pregnancy. Shotgun metagenomic sequencing permitted refinement of the taxonomy established by amplicon sequencing, and identification of functional genes. Additionally, we analyzed saliva and urine metabolomes using GC-TOF and LC-MS/MS lipidomics approaches for samples from mothers and their infants through the first year of life. ResultsAmplicon sequencing revealed most women had either a simple community with one highly abundant species of Lactobacillus or a more diverse community characterized by a high abundance of Gardnerella, as has also been previously described in several independent cohorts. Integrating GC-TOF and lipidomics data with amplicon sequencing, we found metabolites that distinctly associate with particular communities. For example, cervicovaginal microbial communities dominated by Lactobacillus crispatus also have high mannitol levels, which contradicts the basic characterization of L. crispatus as a homofermentative Lactobacillus species. It may be that fluctuations in which Lactobacillus dominate a particular vaginal microbiome are dictated by the availability of host sugars, such as fructose, which is the most likely substrate being converted to mannitol. Furthermore, indole-3-lactate (ILA) was also indicative of L. crispatus specifically. ILA has immunomodulatory properties through binding the human aryl hydrocarbon receptor (AhR), which may maintain the especially low diversity of L. crispatus dominated communities. ConclusionsOverall, using a multi- omic approach, we begin to address the genetic and molecular means by which a particular vaginal microbiome becomes vulnerable to large changes in composition.

microbiology↗