This case study follows a similar approach to the Multiblock sPLS Gastrulation Case Study but includes all five datasets from the single-cell data instead of just three. The goal is to show what can happen when datasets that aren’t well-matched are combined. Figure 1 (reused from the original case study) shows that the RNA data can separate some cell types. However, when looking at the gene body and promoter accessibility datasets (Figures 2 and 3), the sample groupings disappear, and the components become much less informative. This suggests that the datasets do not share enough signal for multiblock sPLS to work well. It highlights the importance of checking each dataset’s output and, when integration fails, trying simpler pairwise comparisons with spls() instead.
📄 Download R script
Data used on this page:
External Gastrulation dataset