Muniak, M. | Search bioRxiv

bioRxiv2023

Fine-tuning TrailMap: The utility of transfer learning to improve the performance of deep learning in axon segmentation of light-sheet microscopy images

Light-sheet microscopy has made possible the 3D imaging of both fixed and live biological tissue, with samples as large as the entire mouse brain. However, segmentation and quantification of that data remains a time-consuming manual undertaking. Machine learning methods promise the possibility of automating this process. This study seeks to advance the performance of prior models through optimizing transfer learning. We fine-tuned the existing TrailMap model using expert-labeled data from noradrenergic axonal structures in the mouse brain. By fine-tuning the final two layers of the neural network at a lower learning rate of the TrailMap model, we demonstrate an improved recall and an occasionally improved adjusted F1- score within our test dataset over using the originally trained TrailMap model. Availability and implementation: The software and data are freely available at https://github.com/pnnl/brain_ohsu and https://data.pnl.gov/group/204/nodes/dataset/35673, respectively.

neuroscience↗

bioRxiv2022

Exogenous sequences in tumors and immune cells (exotic): a tool for estimating the microbe abundances in tumor RNAseq data

The microbiome affects cancer, from carcinogenesis to response to treatments. New evidence suggests that microbes are also present in many tumors, though the scope of how they affect tumor biology and clinical outcomes is unclear. A broad survey of tumor microbiome samples across several independent datasets is needed to identify robust correlations for follow-up testing. We created a tool to carefully identify the tumor microbiome within RNAseq datasets and then applied it to samples collected through the Oncology Research Information Exchange Network (ORIEN) and The Cancer Genome Atlas (TCGA). We showed how the processing removes contaminants and batch effects to yield microbe abundances consistent with non-high-throughput sequencing-based approaches. We sought to establish clinical relevance by correlating the microbe abundances with various clinical and tumor measurements, such as age and tumor hypoxia. This process leveraged the two datasets and raised up only the concordant (significant and in the same direction) associations. We identify associations with survival and clinical variables that are highly cancer-specific and relatively few associations with immune composition. Finally, we explore potential mechanisms by which microbes and tumors may interact using a network approach. Alistipes, a common gut commensal, showed the highest network degree centrality and was associated with genes related to metabolism and inflammation. The exotic tool can support the discovery of microbes in tumors in a way that leverages the many existing and growing RNAseq datasets. Statement of SignificanceThe intrinsic tumor microbiome holds great potential for its ability to predict various aspects of cancer biology and as a target for rational manipulation. Here, we describe a tool to quantify microbes from within tumor RNAseq and apply it to two independent datasets. We show new associations with clinical variables that justify biomarker uses and more experimentation into the mechanisms by which tumor microbiomes affect cancer outcomes.

cancer biology↗