This is a test version of Biostars. For the public version, visit https://www.biostars.org.
Need help in Single cell proteomics data analysis

Hi everyone,

I have single-cell proteomics data from different developmental stages and am exploring analysis options. I'm wondering if it would be appropriate to use scRNA-seq analysis packages, such as Seurat v5, to integrate this data and identify cell cluster markers.

The data was provided by our mass spectrometry platform as a normalized TSV file. The rows represent protein names, and the columns correspond to individual cell names, similar to the format of a typical gene expression matrix.

Here's a small example of what the table looks like:

enter image description here

single cell
• 8,099 views • link
updated by Gordon Smyth Gordon Smyth • written by Saber_J • Saber_J

Yes, I've already looked into this workflow. Since my data is already normalized, I was wondering if I could directly use the Seurat package. However, I've noticed that the scp package also provides functions for batch correction, dimension reduction, and useful visualization tools.

• 0 views • link
written by Saber_J • Saber_J

Probably that's fine but be sure to skip the normalization steps. Also avoid count-based steps such as sctransform for feature selection. Should all be done directly on this matrix which is log2 I suppose. That data matrix (probably you want imputation) needs to go into right slot of the log2-transformed counts. From there on it should be more or less the same.

• 0 views • link
written by ATpoint ATpoint

Thanks for the reply and warm reminder. This table also provides count-like data without normalization, where the values are integers. If I want to use the sctransform method for dimensionality reduction, can I use this integer count data? However, our technician suggested I directly use the normalized data for analysis, as it is a result calculated directly from the commercial software we purchased.enter image description here

• 0 views • link
written by Saber_J • Saber_J

I was assuming your table is log2 intensity as often for MS-based proteomics. I do not know this sort of data you have so I cannot comment.

• 0 views • link
written by ATpoint ATpoint

1 answer

See the single cell proteomics analysis presented in this preprint:

Li M, Cobbold SA, Smyth GK (2025). Quantification and differential analysis of mass spectrometry proteomics data with probabilistic recovery of information from missing values. bioRxiv https://doi.org/10.1101/2025.04.28.651125.

I don't view the Seurat or scp pipelines as adequate, because single proteomics has a high proportion of missing values. The missing values are not missing at random, and need to be handled in a more sophisticated way than ignoring them or routine imputation. Missing values in mass spectrometry data are quite different from zeros in RNA-seq data.

• 0 views • link
written by Gordon Smyth Gordon Smyth

Log in to answer this question.

More posts like this