Quantification

Generally in bottom-up LC-MS/MS proteomics, you quantify precursors, which are broken up pieces of a protein (peptide) that contain some charge state. To make biological sense of quantified signal, it is useful to combine these precursors into their parent protein for downstream analysis. This process of protein quantification can be tricky, as there is no set standard that should always be used.

Tip

Our research tends to focus on using DIA-MS, and we restrict the spectral libraries used to analyze the data to proteotypic peptides, meaning that each precursor in the library is only linked to 1 protein. This makes protein quantification easier as no assumptions need to be made about precursors shared between proteins.

DPKS provides 2 main protein quantification methods:

  1. Absolute Quantification: Using the top_n method.
  2. Relative Quantification: Using the maxlfq method.

Absolute Quantification

Absolute quantification is performed using the top_n method by combining a specified number of the most abundant precursors for each protein using a summarization method (sum, mean, or median).

Note

This is particularly useful if you want to compare proteins to other proteins in an experiment, like with a protein rank plot, to see what proteins are most abundant in your samples.

Absolute quantification can be performed as follows:

qm = qm.quantify(
    method="top_n",
    top_n=3
)

The top_n parameter indicates how many of the precursors you want to use per protein for quantification.

Relative Quantification

DPKS uses the iq1 implementation of the MaxLFQ algorithm2 to extract optimal ratios between samples for each protein and combines them into a resulting protein quantity.

Note

Since this relative quantification approach uses all precursors for a protein, this is not suitable for protein rank plots, as certain proteins will have their absolute abundance underestimated. It is, however, considered state-of-the-art when measuring the differences in protein abundance between 2 experimental groups.

Relative quantification can be performed as follows:

qm = qm.quantify(
    method="maxlfq",
    level="protein",
    threads=5
)

Relative quantification takes much longer to process than absolute quantification, so we have optimized for performance using Numba for JIT compilation of the relative quantification algorithms. To enable multithreading, indicate the desired number of threads using the threads parameter.

The level parameter indicates what level you want to quantify, whether it is proteins, peptides, or precursors. It is possible to quantify peptides using precursors, and precursors using fragment quantities (for DIA experiments) if the correct columns are supplied to the QuantMatrix.

Combined Quantification

To get the best of both worlds, a combined quantification approach is possible that applies relative quantification to the indicated top_n percursors for each protein. This allows for the overall abundance rank of the protein to remain intact while reaping the benefits of signal smoothing and ratio extraction used for relative quantification.

Combined quantification can be performed simply by providing a value >0 to the top_n parameter of the quantify() method:

qm = qm.quantify(
    method="maxlfq",
    level="protein",
    threads=5,
    top_n=3
)

  1. Thang V Pham, Alex A Henneman, Connie R Jimenez, iq: an R package to estimate relative protein abundances from ion quantification in DIA-MS-based proteomics, Bioinformatics, Volume 36, Issue 8, April 2020, Pages 2611–2613, https://doi.org/10.1093/bioinformatics/btz961 ↩

  2. Cox, Jürgen et al. Accurate Proteome-wide Label-free Quantification by Delayed Normalization and Maximal Peptide Ratio Extraction, Termed MaxLFQ. Molecular & Cellular Proteomics, Volume 13, Issue 9, 2513 - 2526, https://doi.org/10.1074/mcp.M113.031591 ↩