The Black Death, Epigenomic MERFISH, Single-Cell Long-Read Sequencing, Be Mindful of the Defaults, and more

2022-W42
natural selection
epigenomics
annotation
MAS-Seq
Author

Mikhael D. Manurung

Published

October 24, 2022

Applications

The Black Death left its mark not only on our history but also on our immunity. Luis Barreiro’s lab showed that natural selection driven by the Black Death, that were protective at that time, could be linked to present-day susceptibility to autoimmune diseases, such as Crohn’s disease.1 Great combinations of ancient DNA work, population genomics, and in-vitro infections.

Nicolet and Wolkers showed that mRNA expression levels are poorly correlated with protein abundance (Pearson r ~ 0.4).2 Interestingly, certain gene charactistics, such as untranslated region (UTR) lengths, could affect mRNA-protein correlation.

Figure 1E: Correlation between mRNA and protein abundance in activated T cells.

Methods

Cell-type annotation is a critical step in single-cell analysis pipeline. For scATAC-Seq, we can annotate cells using either tools developed for scRNA-Seq or transfer labels from RNA to ATAC. Here, Yuge Wang and colleagues performed a thorough benchmarking and showed subpar performance of all these methods.3 Using Multiome data as reference can help improve the accuracy, but only to some extent.

Weighted accuracy of cell-type prediction. Note the <50% accuracy for all methods.

Technologies

Xiaowei Zhuang published Epigenomic MERFISH, an imaging-based technology for spatially resolved single-cell epigenomic profiling.4 Epigenomic MERFISH maps active promoters and enhancers in mouse brain and reveals putative enhancer hubs in embryonic mouse brain.

Awesome spatial technologies seem to be published on a weekly basis. I wonder which ones will be commercialized and used around the world?

Epigenomic MERFISH

PacBio, in partnership with the Broad Institute and 10X Genomics, launched the new Multiplexed Arrays Sequencing (MAS-Seq).5 This technology enables long-read scRNA-Seq for identification of alternatively spliced isoforms and gene fusions without the need for de novo assembly.6

Excellent move by 10X’s single-cell division! Looking forward to the maneuvers of its spatial division.

Conversations

Inter-operability of single-cell analysis frameworks is not enough, the tools should be coherent as well! Log-fold change calculation in Seurat and Scanpy differs significantly7, as pointed out by Nikolai Slavov from a recent benchmarking paper by Davis McCarthy’s lab.8 This mess begs for a question: to whom should we handle the responsibility for making the decisions in single-cell data analysis? Developers or Users?

Related to the post above, Nicola Romanò shared his observation on the danger of default parameters in single-cell analysis frameworks.9 If you’re using Seurat, remember to pass your seed into its function instead of declaring it before calling the function, as per usual, or else the function will override your seed.

Yet another thread on the danger of (blindly following) the defaults, Junhyong Kim shared his thoughts on Seurat’s CP10K normalization (i.e. dividing gene count by library depth, multiply by 10K, followed with log1p).10. Related to this, Lior Pachter’s lab published a benchmark on normalisation methods and suggested proportional fitting –> log-transformation –> proportional fitting (PFlog1pPF).11 Or should we abandon the normalisation route and instead fully embrace the count nature of scRNA-Seq data?12

Getting confused now? Remember to include bioinformatician(s) in your team from the start.

Miscellaneous

Jobs

Luciano Martelotto has an open Senior Scientist position.


Stay in touch

If you enjoyed this post, then don't miss out on any future posts by subscribing to my newsletter.

Share