Biology subjects

McMillan, C.

Publications and source records attributed to McMillan, C..

7 recordsLinked to original sources

bioRxiv2025

Integrating 12 Spatial and Single Cell Technologies to Characterise Tumour Neighbourhoods and Cellular Interactions in three Skin Cancer Types

Cutaneous squamous cell carcinoma (cSCC), basal cell carcinoma (BCC), and melanoma - the three major skin cancers - collectively comprise over 70% of all cancer cases. Despite their prevalence, much understanding of cellular interactions in the skin cancer microenvironment is needed, both in the outer skin layer where the cancer originates and at the deeper junctional and dermal layers into which it progresses. To address this gap, we integrated 12 complementary spatial and single-cell technologies to generate orthogonally-validated cell signatures, spatial maps, and interactomes for cSCC, BCC, and melanoma. Through comprehensive comparisons and integrating these spatial methods, we provided practical benchmarking guidelines for experimental design and analysis. By identifying keratinocyte cancer cells and melanomas, we found distinct signatures of these cells compared to non-cancer keratinocytes and melanocytes. Spatial integration of transcriptomics, proteomics and glycomics uncovered cancer niches enriched for cancer initiating cells (melanocytes or keratinocytes) and fibroblast and T-cell (MKFT) clusters, with altered tyrosine and pyrimidine metabolism. Ligand-receptor analysis across >700 cell-type combinations and >1.5 million interactions highlighted key roles for CD44, integrins, and collagens, with CD44-FGF2 emerging as a potential therapeutic target. Consistently, melanoma showed strong MFT interactions, validated by Opal Polaris, RNAScope, Proximal Ligation Assay and two additional single-cell spatial platforms (making a total of 14 technologies). For population-scale generalisation, genetic associations from >500,000 individuals were mapped onto spatial skin tissues, identifying SNPs enriched in domains containing melanocytes and T cells and their ligand-receptor pairs, shedding light on functional mechanisms linking genetic heritability to cells within cancer tissue. We built an interactive multiomics resource for exploring spatially-resolved molecular signatures and cellular crosstalk in skin cancer, available at https://skincanceratlas.com.

cancer biology↗

bioRxiv2025

Parameter Efficient Fine-tuning of Transformer-based Masked Autoencoder Enhances Resource Constrained Neuroimage Analysis

Recent innovations in artificial intelligence (AI) have increasingly focused on large-scale foundational models that are more general purpose in contrast to conventional models trained to perform specialized tasks. Transformer-based architectures have become the standard backbone in foundation models across data modalities (image, text, audio, video). There has been a keen interest in applying parameter-efficient fine-tuning (PEFT) methods to adapt these models to specialized downstream tasks in language and vision. These methods are particularly essential for medical image analysis where the limited availability of training data could lead to overfitting. In this work, we evaluated different types of PEFT methods on pre-trained vision transformers relative to typical training approaches, such as full fine-tuning and training from scratch. We used a transformer-based masked autoencoder (MAE) framework, to pretrain a vision encoder on T1- weighted (T1-w) brain MRIs. The pretrained vision transformers were then fine-tuned using different PEFT methods that reduced the trainable model parameters to as few as 0.04% of the original model size. Our study shows that: 1. PEFT methods were competitive with or outperformed the reference full fine-tuning approach and outperformed training from scratch, with only a fraction of the trainable parameters; 2. PEFT methods with a 32% reduction in model size boosted Alzheimers disease (AD) classification by 3% relative to full fine-tuning and 11% relative to a 3D CNN, with only 258 training scans; and 3. PEFT methods performed well on diverse neuroimaging tasks including AD and Parkinsons disease (PD) classification, and "brain-age" prediction based on T1-w MRI datasets - a standard benchmark for deep learning models in neuroimaging; 4. smaller model sizes were competitive with larger models in test performance. Our results show the value of adapting foundation models to neuroimaging tasks efficiently and effectively in contrast to training stand- alone special purpose models.

neuroscience↗

bioRxiv2025

Developing an anatomically valid segmentation protocol for anterior regions of the medial temporal lobe for neurodegenerative diseases

BackgroundThe anterior portion of the medial temporal lobe (MTL) is one of the first regions targeted by pathology in sporadic Alzheimers disease (AD) and Limbic-predominant Age-related TDP-43 Encephalopathy (LATE) indicating a potential for metrics from this region to serve as imaging biomarkers. Leveraging a unique post-mortem dataset of histology and magnetic resonance imaging (MRI) scans we aimed to 1) develop an anatomically valid segmentation protocol for anterior entorhinal cortex (ERC), Brodmann Area (BA) 35, and BA36 for in vivo 3 tesla (T) MRI and 2) incorporate this protocol in an automated approach. MethodsWe included 20 cases (61-97 years old, 50% females) with and without neurodegenerative diseases (11 vs. 9 cases) to ensure generalizability of the developed protocol. Digitized MTL Nissl-stained coronal histology sections from these cases were annotated and registered to same-subject post-mortem MRI. The protocol was developed by determining the location of histological borders of the MTL cortices in relation to anatomical landmarks. Subsequently the protocol was applied to 15 cases twice, with a 2-week interval, to assess intra-rater reliability with the Dice Similarity Index (DSI). Thereafter it was implemented in our in- house Automatic Segmentation of Hippocampal Subfields (ASHS)-T1 approach and evaluated with DSIs. ResultsThe anterior histological border distances of ERC, BA35 and BA36 were evaluated with respect to various anatomical landmarks and the distance relative to the beginning of the hippocampus was chosen. To formulate segmentation rules, we examined the histological sections for the location of borders in relationship to anatomical landmarks in the coronal sections. The DSI for the anterior MTL cortices for the intra-rater reliability was 0.85-0.88 and for the ASHS-T1 against the manual segmentation was 0.62-0.65. DiscussionWe developed a reliable segmentation protocol and incorporated it in an automated approach. Given the vulnerability of the anterior MTL cortices to tau deposition in AD and LATE, the updated approach is expected to improve imaging biomarkers for these diseases.

neuroscience↗

bioRxiv2024

Cytoarchitectonic gradients of laminar degeneration in behavioral variant frontotemporal dementia

Behavioral variant frontotemporal dementia (bvFTD) is a clinical syndrome primarily caused by either tau (bvFTD-tau) or TDP-43 (bvFTD-TDP) proteinopathies. We previously found lower cortical layers and dorsolateral regions accumulate greater tau than TDP-43 pathology; however, patterns of laminar neurodegeneration across diverse cytoarchitecture in bvFTD is understudied. We hypothesized that bvFTD-tau and bvFTD-TDP have distinct laminar distributions of pyramidal neurodegeneration along cortical gradients, a topologic order of cytoarchitectonic subregions based on increasing pyramidal density and laminar differentiation. Here, we tested this hypothesis in a frontal cortical gradient consisting of five cytoarchitectonic types (i.e., periallocortex, agranular mesocortex, dysgranular mesocortex, eulaminate-I isocortex, eulaminate-II isocortex) spanning anterior cingulate, paracingulate, orbitofrontal, and mid-frontal gyri in bvFTD-tau (n=27), bvFTD-TDP (n=47), and healthy controls (HC; n=32). We immunostained all tissue for total neurons (NeuN; neuronal-nuclear protein) and pyramidal neurons (SMI32; non-phosphorylated neurofilament) and digitally quantified NeuN-immunoreactivity (ir) and SMI32-ir in supragranular II-III, infragranular V-VI, and all I-VI layers in each cytoarchitectonic type. We used linear mixed-effects models adjusted for demographic and biologic variables to compare SMI32-ir between groups and examine relationships with the cortical gradient, long-range pathways, and clinical symptoms. We found regional and laminar distributions of SMI32-ir expected for HC, validating our measures within the cortical gradient framework. While SMI32-ir loss was not related to the cortical gradient in bvFTD-TDP, SMI32-ir progressively decreased along the cortical gradient of bvFTD-tau and included greater SMI32-ir loss in supragranular eulaminate-II isocortex in bvFTD-tau vs bvFTD-TDP (p=0.039). In a structural model for long-range laminar connectivity between infragranular mesocortex and supragranular isocortex, we found a larger laminar ratio of mesocortex-to-isocortex SMI32-ir in bvFTD-tau vs bvFTD-TDP (p=0.019), suggesting select long-projecting pathways may contribute to isocortical-predominant degeneration in bvFTD-tau. In cytoarchitectonic types with the highest NeuN-ir, we found lower SMI32-ir in bvFTD-tau vs bvFTD-TDP (p=0.047), suggesting pyramidal neurodegeneration may occur earlier in bvFTD-tau. Lastly, we found that reduced SMI32-ir related to behavioral severity and frontal-mediated letter fluency, not temporal-mediated confrontation naming, demonstrating the clinical relevance and specificity of frontal pyramidal neurodegeneration to bvFTD-related symptoms. Our data suggest loss of neurofilament-rich pyramidal neurons is a clinically relevant feature of bvFTD that selectively worsens along a frontal cortical gradient in bvFTD-tau, not bvFTD-TDP. Therefore, tau-mediated degeneration may preferentially involve pyramidal-rich layers that connect more distant cytoarchitectonic types. Moreover, the hierarchical arrangement of cytoarchitecture along cortical gradients may be an important neuroanatomical framework for identifying which types of cells and pathways are differentially involved between proteinopathies.

neuroscience↗

bioRxiv2023

Comparison of Anatomical and Diffusion MRI for detecting Parkinson's Disease using Deep Convolutional Neural Network

Parkinsons disease (PD) is a progressive neurodegenerative disease that affects over 10 million people worldwide. Brain atrophy and microstructural abnormalities tend to be more subtle in PD than in other age-related conditions such as Alzheimers disease, so there is interest in how well machine learning methods can detect PD in radiological scans. Deep learning models based on convolutional neural networks (CNNs) can automatically distil diagnostically useful features from raw MRI scans, but most CNN-based deep learning models have only been tested on T1-weighted brain MRI. Here we examine the added value of diffusion-weighted MRI (dMRI) - a variant of MRI, sensitive to microstructural tissue properties - as an additional input in CNN-based models for PD classification. Our evaluations used data from 3 separate cohorts - from Chang Gung University, the University of Pennsylvania, and the PPMI dataset. We trained CNNs on various combinations of these cohorts to find the best predictive model. Although tests on more diverse data are warranted, deep-learned models from dMRI show promise for PD classification. Clinical RelevanceThis study supports the use of diffusion-weighted images as an alternative to anatomical images for AI-based detection of Parkinsons disease.

neuroscience↗

bioRxiv2021

3D Convolutional Neural Networks for Classification of Alzheimer's and Parkinson's Disease with T1-Weighted Brain MRI

Parkinsons disease (PD) and Alzheimers disease (AD) are progressive neurodegenerative disorders that affect millions of people worldwide. In this work, we propose a deep learning approach to classify these diseases based on 3D T1-weighted brain MRI. We analyzed several datasets including the Parkinsons Progression Markers Initiative (PPMI), an independent dataset from the University of Pennsylvania School of Medicine (UPenn), the Alzheimers Disease Neuroimaging Initiative (ADNI), and the Open Access Series of Imaging Studies (OASIS) dataset. The UPenn and OASIS datasets were used as independent test sets to evaluate the model performance during inference. We also implemented a random forest classifier as a baseline model by extracting key radiomics features from the same T1-weighted MRI scans. The proposed 3D convolutional neural network (CNN) model was trained from scratch for the classification tasks. For AD classification, the 3D CNN model achieved an ROC-AUC of 0.878 on the ADNI test set and an average ROC-AUC of 0.789 on the OASIS dataset. For PD classification, the proposed 3D CNN model achieved an ROC-AUC of 0.667 on the PPMI test set and an average ROC-AUC of 0.743 on the UPenn dataset. Model performance was largely maintained when using only 25% of the training dataset. The 3D CNN outperformed the random forest classifier for both the PD and AD tasks. The 3D CNN also generalized better on unseen MRI data from different imaging centers. These approaches show promise for screening of PD and AD patients using only T1-weighted brain MRI, which is relatively widely available. This model with additional validation could also be used to help differentiate between challenging cases of AD and PD when they present with similarly subtle motor and non-motor symptoms.

bioinformatics↗

bioRxiv2021

A SARS-CoV-2 targeted siRNA-nanoparticle therapy for COVID-19

Coronavirus disease 2019 (COVID-19) is caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection in humans. Despite several emerging vaccines, there remains no verifiable therapeutic targeted specifically to the virus. Here we present a highly effective siRNA therapeutic against SARS-CoV-2 infection using a novel lipid nanoparticle delivery system. Multiple small-interfering RNAs (siRNAs) targeting highly conserved regions of the SARS-CoV-2 virus were screened and three candidate siRNAs emerged that effectively inhibit virus by greater than 90% either alone or in combination with one another. We simultaneously developed and screened two novel lipid nanoparticle formulations for the delivery of these candidate siRNA therapeutics to the lungs, an organ that incurs immense damage during SARS-CoV-2 infection. Encapsulation of siRNAs in these LNPs followed by in vivo injection demonstrated robust repression of virus in the lungs and a pronounced survival advantage to the treated mice. Our LNP-siRNA approaches are scalable and can be administered upon the first sign of SARS-CoV-2 infection in humans. We suggest that an siRNA-LNP therapeutic approach could prove highly useful in treating COVID-19 disease as an adjunctive therapy to current vaccine strategies.

synthetic biology↗