Early vs. late timepoints

My PLSDA and ASCA analyses showed a strong separation between early and late timepoints for metabolomes at both temperatures. Based on this, I decided to go back and repeat the PCA, perMANOVA, ASCA, and PLSDA analyses using the early vs. late distinction.

• PCAs found here
• perMANOVA results here, beta dispersion test results here
• ASCA figures here, and loadings here
• PLSDAs here

Overall, I see some good separation between early and late timepoints (which is what I expected). Redoing all of these things was just due diligence. Having the PLS-DA split by early and late timepoints also allows me to pull out some higher-order VIPs. I will spend some time talking about the ASCA output because I think that is the most different.

Figures 1-3. Effect plots for significant PCs from timepoint ASCA

PC 1 shows metabolites with higher abundance at 25ºC, but that decrease over time across both temperatures. This is pretty consistent with what I saw when considering all timepoints! PC 2 shows metabolites with higher abundance at 25ºC that increase over time across both temperatures. The slope, however, is steeper for 25ºC than 5ºC. This is actually much cleaner to interpret when compared to the ASCA done with all timepoints, particularly at 5ºC. The loadings are also stronger (no overlap with 0). PC 3 shows metabolites that increase over time at 5ºC but decrease over time at 25ºC. Even though the loadings aren’t as strong at PC 3 barely explains > 10% of the variation, I think there are some interesting potential trends here.

Going forward

1. VIP analysis for major trends
2. Index, get advanced transcriptome statistics, and pseudoalign with salmon
3. Clean transcriptome with EnTap and blastn
4. Identify differentially expressed genes
5. Additional strand-specific analysis in the supergene region
6. Examine HOBO data from 2023 experiment
7. Demographic data analysis for 2023 paper
8. Start methods and results of 2023 paper