Signature Analysis - PMSB5 and generic resistance signatures

[gwaybio]

In #57 I add a signature analysis. The purpose of this analysis is to identify morphology features that are significantly different between wildtype and resistant clones. The next step is to apply the signatures to other profiles to 1) validate the approach and 2) predict the resistance status of different samples.

This analysis was prompted by #49 .

I will describe the approach, results, and conclusions in this issue.

[gwaybio]

Hypothesis

We can identify morphology features that distinguish resistant from wild-type clones

Data

We use two different (aggregated) datasets

1. Batch 1 (20X) and Batch 2
   • Clone A and E (confirmed PSMB5 mutations)
2. Batch 5, 6, 7
   • "Four Clone" (we have four resistant and four wild-type clones)
   • DMSO and 0.7 Bortezomib (I think it is lowest dose?)

Approach

• Only select untreated samples
• Build a linear model to explore which metadata features contribute to observed variation
• Metadata: Batch (plate), Cell Line, Clone Type
  • Should we include well?
• Select morphology features that have a limited contribution to certain variables
• Remove features that have high batch and inter-individual contributions
• Isolate the effect of Clone Type
• Select features that have significant Clone Type differences by Tukey’s HSD
• Repurpose single-sample gene set enrichment (ssGSEA) methods
  • Gives an enrichment score per profile
  • Expect certain clones to have high scores, others to not

Limitations

• Small data to generate PSMB5 mutations
  • Currently only using untreated samples
  • Perhaps we can increase the amount of data by including low dose in building signatures
• clone A and E only represent PSMB5 mutations
  • Core resistance machinery vs. specific to PSMB5?

[gwaybio]

Signature Results

Linear Model

Clone A/E

Four Clone

[gwaybio]

Signature Results

Applying Signatures

Here, we repurpose singscore, which is a rank-based ssGSEA-like method that generates a composite score of high and low features.

Apply PSMB5 Signature to Four Clone

Apply "Generic" Signature to Clone A/E

[gwaybio]

In general, it appears that the resistance signature decreases with increasing dose. However, this decrease is dependent on clone type (resistant vs. wildtype).

Alternative Plotting Strategy

[shntnu]

Here are some snippets that may be relevant when we ponder the question of of stratifying our pseudo-bulk analysis

From https://osca.bioconductor.org/multi-sample-comparisons.html#differential-expression-between-conditions

• Collapsing cells into samples reflects the fact that our biological replication occurs at the sample level (Lun and Marioni 2017). Each sample is represented no more than once for each condition, avoiding problems from unmodelled correlations between samples. Supplying the per-cell counts directly to a DE analysis pipeline would imply that each cell is an independent biological replicate, which is not true from an experimental perspective. (A mixed effects model can handle this variance structure but involves extra statistical and computational complexity for little benefit, see (???).)
• Variance between cells within each sample is masked, provided it does not affect variance across (replicate) samples. This avoids penalizing DEGs that are not uniformly up- or down-regulated for all cells in all samples of one condition. Masking is generally desirable as DEGs - unlike marker genes - do not need to have low within-sample variance to be interesting, e.g., if the treatment effect is consistent across replicate populations but heterogeneous on a per-cell basis. (Of course, high per-cell variability will still result in weaker DE if it affects the variability across populations, while homogeneous per-cell responses will result in stronger DE due to a larger population-level log-fold change. These effects are also largely desirable.)