boyiguo1/manuscript-spaqc

Manuscript repo for the project of novel spatially-aware quality control methods

manuscript-spaqc's Issues

Literature Review

Summarize commonly used QC metrics in Single Cell data analysis

• Method 1
• Method 2
• Method 3

Recommended QC metrics/method for spatially-resolved data.

Possible demostrations of the improved utility of SpaQC (Brainstorming)

We know that the main competitor methods are a couple of single cell QC measurements/standards. Is there any scenarios that we envision spatially-aware QC metrics possible could advance? What would be some good examples to demonstrate the advantages of applying spatially-aware QC?

Optimally, we would hope that spaQC would outperform all the competitors based on some metrics in various case studies. But it is acceptable (for publishing purpose) that we could that spaQC outperforms each competing methods in individual case studies. That would be the evdience to support a claim like spaQC would be the most comprehensive QC method.

In addition, if we can think of/find cases that different QC metric would have substaintial impact on the downstream analysis, this would be be extremely solid evidence.

Methods of comparison

• Method 1, e.g. fixed thresholding
  • Not data driven
• Method 2, e.g. Median Absolute Deviation
  • Would remove rare/extreme cell population/spatial domain
• (optional) Method 3

Case study:

• Case 1 - Utility 1

• Case 2 - Utility 2

  • data-driven QC instead of heuristic driven QC
    • Don't have to use arbitrary threshold for all samples, or manually select for each sample.
  • ...

  This a brainstorming post. I would love to see how you would think about this. Or just keep this at the back of your head when explore different methods.

Evidence for scQC fails

• Visualization

  • Thresholding can systematically remove cell population, visualizing via spatial plots and matched anatomical regions.

• (Optional) Theoretical Evidence

  • Threshold mitochondria reads confounds with biology due to change in experimental protocal

Potential solutions

• Neighborhood outlier

  • A function identify K-nearest neighbourhood based on physical distance instead of eucliean distance

    • parameters:
      • K nearest neighbours
      • Maximum searching distance for the nearest neighbors
    • Bioc Function to find nearest neighbors: https://www.bioconductor.org/packages/release/bioc/html/BiocNeighbors.html

  • Small-sample outlier detection method

• How do we define "hangnail" regions?

The objective of spatially aware QC

@MicTott Do you have any vision or preferred definition of spatially-aware QC? For example, do you expect the QC happens at only the spot level, or as well as some micro-envrionment level (e.g. identify "hangnail" regions).

We don't have to them all depending on how challenging they are, but we could/should establish some priority that ouwld help us to define the tentative scope of the project.

Thanks