Session 4

Multiplexed Ion Beam Imaging (MIBI) analysis pipeline: in situ characterisation of tissue architecture by open-source analysis

Mrs Nina Tubau Ribera, Bioimage analyst, WEHI

Co-authors
Ms Claire Marceaux, WEHI
Dr Kelly Rogers, WEHI
Dr Marie-Liesse Asselin-Labat, WEHI
Dr Lachlan Whitehead, WEHI

Multiplexed Ion Beam Imaging (MIBI) is a recent technology enabling comprehensive phenotypic profiling and spatial analysis of solid tissue. It uses a dynamic secondary ion mass spectrometer instrument with a time-of-flight mass analyser to image antibodies tagged with monoisotopic metal reporters and map their location in a multi-channel image. The system allows to visualise up to 40 markers on paraffin-embedded or fresh-frozen samples at subcellular resolution. This platform is available at Walter and Eliza Hall Institute of Medical Research since January. The system generates multi-layer TIFF images that are supported by Bio-Formats and are analysed in our recently developed pipeline using the open-source software QuPath.

First, we implemented a tiling and stitching pipeline, so users can easily image multiple fields of view of larger tissue sections. The image preprocessing consists of an isobaric correction and Voronoi tesselation filtering. A further step of normalisation has been added in order to complete any meaningful quantification. Within QuPath, we have implemented a segmentation tool using Deep Cell machine learning segmentation that provides state-of-the-art cell detection. In addition to the segmentation, we added a tool to measure the cell positivity to each marker based on mean intensity throughout the previously segmented cell. From the cell phenotypes we directly perform spatial analysis and visualise the results. Our current work resides in joining all these steps in a single-button pipeline allowing users an easy, quick and reproducible workflow for highly multiplexed analysis.

Automated annotation and visualisation of high-resolution spatial proteomic mass spectrometry imaging data using HIT-MAP

A/Prof Thomas Cox, Laboratory Head, Garvan Institute of Medical Research

Biological tissues are highly compartmentalized due to their complex and diverse functions. Organs and tissues are partitioned into histologically distinct, yet functionally co-dependent sub-regions that exhibit diverse cellular and molecular compositions. Importantly, the unique set of expressed proteins, specific to a particular cell type, location, or place in time and space, critically underpins tissue and organ function. Significant changes in these proteomes are observed in almost all disease-states.

Mass Spec Imaging has the potential to significantly advance our understanding of biology, physiology and medicine. Matrix-assisted laser desorption/ionisation mass spectrometry imaging (MALDI-MSI) is a powerful tool in the spatial proteomics field, enabling direct detection and registration of protein abundance and distribution across tissues. MALDI-MSI preserves spatial distribution and histology allowing unbiased analysis of complex, heterogeneous tissues.

However, MALDI-MSI faces the challenge of simultaneous peptide quantification and identification. To overcome this, we developed and validated HIT-MAP (High-resolution Informatics Toolbox in MALDI-MSI Proteomics), an open-source bioinformatics workflow using peptide mass fingerprint analysis and a dual scoring system to computationally assign peptide and protein annotations to high mass resolution MSI datasets and generate customisable spatial distribution maps.

The unbiased nature of MALDI-MSI allows for the interrogation of whole proteomes. Furthermore, the integration and co-registration of MALDI-MSI datasets with other established and/or emerging technology platforms (such as histology and spatial transcriptomics) will significantly increase our understanding of health and disease through combining complementary orthogonal data types.

HIT-MAP will be a valuable resource for the spatial proteomics community for analysing newly generated and retrospective datasets in both normal and disease contexts.

SNT: a unifying toolbox for quantitative neuroanatomy

Dr Tiago Ferreira, Howard Hughes Medical Institute Janelia Research Campus

Multi-dimensional and multi-modal imagery are becoming commonplace in cellular neuroscience. Historically, software for neuronal reconstruction has not been amenable to such datasets. To this end, we have developed SNT, an end-to-end framework for neuronal morphometry that supports tracing, proof-editing, visualization, quantification, and modeling of neuroanatomy with support for whole-brain projectomes. With an open architecture, a large user base, community-based documentation, support for complex imagery and several model organisms, SNT became a broad resource for the broad neuroscience community. SNT is both a desktop application and multi-language scripting library, and it is available through the Fiji distribution of ImageJ.

Computational 3D fluorescence microscopy

Associate Professor Laura Waller, University of California Berkley

We describe a compact and inexpensive computational microscope that encodes 3D information into a single 2D sensor measurement, then exploits sparsity to reconstruct the volume with good resolution across a large volume.  Our system uses simple hardware and scalable software for easy reproducibility and adoption. The inverse algorithm is based on large-scale nonlinear optimization with self-calibration of aberrations and we discuss computational optical design approaches for optimizing the system’s performance. We demonstrate applications in whole organism bioimaging and neural activity tracking in vivo.