In our recent Akoya Academy webinar, “A Deep Dive on the Multiplex Analysis Viewer (MAV) to Analyze CODEX Images”, we heard from Grady Carlson, PhD, Technical Application Scientist at Akoya Biosciences. Grady introduced the CODEX MAV software and its features, including gating, unsupervised clustering, dimensionality reduction plots (e.g., tSNE), and spatial interaction plots and tables.
A t-SNE plot in the CODEX MAV software
We received several questions during our Q&A session after the talk. Since we weren’t able to get to all of them during the session, Grady has answered them here.
Did you miss the live webinar? Watch it on-demand.
How do we install and update MAV?
Installation instructions are available here. To update CODEX MAV, please update the FIJI plugins by selecting “Help” and “Update” from the main FIJI menu. CODEX MAV will update automatically with other FIJI plugins.
How are cells determined and segmented?
Presently, cells are segmented by the CODEX processor. The processor uses a Watershed algorithm on DAPI or Hoechst positive pixels. Size and intensity cutoffs are user-specified and recommended settings are available here. Custom segmentation is also possible. Changes will be coming soon related to cell segmentation.
What is the significance of using DAPI-2 as opposed to DAPI-1 in analysis?
The DAPI-1 signal is acquired during the first cycle and is thus the product of manual staining. The DAPI signal is therefore almost always more intense than DAPI signals in other cycles. Thus DAPI-2, 3, 4,… are representative of the DAPI signals throughout the CODEX experiment and are used as the input channels for nuclear cell segmentation.
How does MAV manage tissue microarrays (TMA)? How do you segment each TMA core?
Currently, MAV treats each core of a TMA as a separate region. TMA cores can easily be combined using gating or kept separate for analysis. All cores are segmented identically using the parameters specified in the CODEX processor.
Can CODEX MAV be used on any multiplex images or only images using CODEX?
CODEX MAV was designed to support images produced by CODEX. The use of CODEX MAV to analyze images generated on a non-Akoya system is not supported.
Does the size of the tissue or region affect the analysis speed?
The answer to this question is largely dependent on the computer used to analyze CODEX imagery. If the recommended CODEX analysis computer is used to analyze images, changes in analysis speed due to region size will be minimal and likely unnoticeable. However, on a computer with less RAM, larger tissues or regions could lead to slower analysis due to limited RAM.
What does interaction mean in Spatial Analysis?
In CODEX MAV, we define interaction by spatial proximity. CODEX MAV allows the user to define the distance used to calculate the number of interacting cells; an interacting cell will be any cell that falls within a predefined spatial distance range (e.g., 5-10 μm).
Is it possible to get different clustering results when the same parameters are used at two different time points?
Yes, it is possible to see a difference in the number of clusters that are produced by the algorithm. The reason for this is that every time you run the clustering algorithm, different seed points are selected, and this may result in different cluster numbers.
Is it possible to modify parameters that will affect t-SNE results?
Not at present, but we are planning to implement such features.
Is it possible to modify the text size in the Circos plot, which displays spatial relationships?
Yes, directions to change font size are available here.
Can alternative segmentation programs be used other than the CODEX processor?
Yes, however MAV looks for specific column names when importing cells. Please contact us at email@example.com for additional information regarding custom segmentation.
Can Vectra images be analyzed with MAV?
Not at present.
When combining clustered populations are there any good rules of thumb for deciding whether the populations are significantly different or is this determined on a subjective basis.
In the webinar, no statistical significance was shown. In order to establish such a rule of thumb, robust controls for staining would need to be implemented to establish cutoffs for specific markers in control tissue and the tissue of interest. However, for anyone seeking formalized rules of thumb for clustering high-plex data, we recommend checking out this article.
What is the main difference between Opal and CODEX?
Opal and CODEX operate on different chemistries and workflows. In the CODEX workflow, all antibodies are stained simultaneously. During the CODEX experiment, oligo-tagged antibodies are detected by complementary oligo-tagged, fluorescently labeled CODEX Reporters. These reporters are cyclically added, revealed and removed under isothermal conditions using the CODEX system. Every cycle, up to 3 reporters (plus DAPI) are imaged using the integrated companion microscope. Bonds between complementary oligos are hydrogen bonds and are reversible.
Opal, on the other hand, requires sequential staining and stripping steps. Opal utilizes fluorescently labeled Tyramide and, like CODEX, can be used with any primary antibody. In the presence of HRP (catalyst, bound to secondary antibody) and hydrogen peroxide, Tyramide covalently (irreversibly) binds to target proteins in the tissue, localized at the primary antibody. Primary and secondary antibodies are removed from the tissue with the application of heat and the process is repeated until all antigens are detected with Opal.
Interested in learning what else MAV can do? Learn how to perform single-cell analysis of cell suspensions using the CODEX workflow in this recorded webinar.