Low MID counts

[pikapika505]

We have four Arabidopsis Stereo-seq samples, and after running saw count we are seeing very low counts per cell (up to ~350-500 MIDs).

Our first concern is the reference: the saw makeRef pipeline does not appear to accept a GFF3 file as an annotation input, so we used the GTF file instead: Araport11_GTF_genes_transposons.20241001.gtf.gz.

Second, the saw count HTML reports are empty for all samples (e.g. run/Sample8/outs/report/report.html).

Could you please suggest what we should check or adjust in our analysis to troubleshoot these issues?

[Jinghao-Chen]

Hi, actually, saw makeRef supports GFF3 format input. You might want to consider updating your SAW version, as the latest version available is 8.2.1.

If you are using SAW version 8.1.1 or earlier, the html report is provided as a single packaged folder. You will need to download the entire SN.report.tar.gz file to your local machine, extract all its contents into the same directory, and then open report.html in a web browser.

[pikapika505]

@Clouate thank you! Version 8.2.1 worked with the GFF3 file and I now see more genes overall. However, MID counts per cell are still very low in all 4 Arabidopsis leaf samples (median 14–24, mean 18–32).
The only alert in the saw count reports is a very low Fraction MID Spots Under Tissue (~13%).
Do you think this is more likely due to library preparation/sequencing issues, or is there anything else I can try in SAW to improve MID recovery?

[Jinghao-Chen]

Hi, it is suggested that you check if the image and the expression matrix are aligned, and whether the morphology of the tissue expression matrix matches that on the image. Is the tissue area relatively small (refer to the tissue area value in the Tissue section of the html report)? What is the total number of Unique reads in the sun plot? It would be more helpful if an HTML report could be provided.

[pikapika505]

thank you @Clouate ! Here is a report for one sample https://drive.google.com/file/d/1LcgEZuMvQxEiiIQ9yHj4nRv6-C1_At05/view?usp=sharing

Here are some DAPI/ FB28_IF images from Stero-Map

[pikapika505]

also, for one sample out of 4, I got a different format of the report https://drive.google.com/file/d/1dWmop2FE2q8KQinfgQbPYjz9pLl5M3Wm/view?usp=sharing . Do you know why this could happen?

[Jinghao-Chen]

Hi, There are two main causes of the issue:

1. As shown in the figure below, the registration between the expression matrix and the image failed. The tissue orientation in the expression matrix is inconsistent with that in the image, resulting in an incorrect matrix for the tissue region which contains only a minimal number of MIDs.
   It is recommended that you use Stereomap ImageProcessing for manual image processing and execute saw realign to regenerate the results.
   https://stereotoolss-organization.gitbook.io/stereomap-user-manual-v4.2/tutorials/navigation-for-image-processing/nuclei-staining-+-immunofluorescence-image#considerations-for-nuclei-staining-images-and-immunofluorescence-images

2. Additionally, the report indicates a high proportion of multi-mapped reads, leading to a reduced amount of confidently mapped data and ultimately fewer MIDs.
   Regarding the incorrect format of one of the reports, it was likely generated using SAW version 8.1.1. You need to download the entire report.tar.gz file first, then extract it completely before opening.

[pikapika505]

@Clouate thank you for pointing that out. I followed the Stereomap Image Processing instructions and loaded the saw count outputs from run/Sample6/outs/visualization/C05695C5_SC_20250826_124028_4.1.2.tar.gz. The DAPI image aligned with run/Sample6/outs/visualization/C05695C5.stereo, and the tracklines matched as well.

However, in the Feature Point section the image appeared flipped and rotated. Because of that, I selected the “left bottom” point according to the visible mirroring and rotation, but doing so actually misaligned the DAPI image and the expression map. After that, regardless of which point I chose, the images would not align properly again.

If I skip the Feature Point section and register the image as-is, then run saw realign, I end up with the same misalignment that I originally had in the saw count results.

Could you please help me resolve this? Thank you!

[Jinghao-Chen]

Hi, due to the issue of the tissue shape, registration is not easy. In fact, from the expression matrix heatmap, the tissue orientation should be vertical rather than horizontal. Could you share the files outs/feature_expression/C05695C5.raw.gef and C05695C5_SC_20250826_124028_4.1.2.tar.gz with me? I can assist with manual processing. Please send them to my email: chenjinghao@stomics.tech.

Alternatively, you could follow these steps:

1. Run the command: saw convert visualization --gef C05695C5.raw.gef --visualization-gef C05695C5.tissue.gef
   Note: Back up the original C05695C5.tissue.gef first to avoid overwriting with the same filename.
2. Download the newly generated C05695C5.tissue.gef to your Windows computer, and place it in the extracted visualization directory (in the same folder as the .stereo file).
3. Re-run StereoMap ImageProcessing. At Step 2, you will now see the full-chip expression matrix heatmap. This will help you visualize the complete morphology of the tissue on the chip, facilitating morphological registration operations.