Improved performance for analysis of high resolution LC-MS data

Analytica 2014 saw the official launch of the new and rebranded releases of both Progenesis QI and Progenesis QI for proteomics. One of the new features shared by both products is the ability to control the noise reduction during data import – something we’ll look at in more detail here.

Big data just keeps getting bigger

One of the challenges with LC-MS data is the sheer size of the files and the demand this makes on your PC’s resources, specifically its RAM. While Progenesis is well-known for its ability to reduce this burden through its use of an intelligent peak-modelling algorithm, the advancement of PC hardware isn’t keeping pace with our demand for higher resolution instruments and the size of data they produce. Analysing such high-resolution datasets can still pose a significant challenge, but Progenesis holds an ace card in the guise of its filter strength setting:

Filter strength dialog

This setting provides the option of applying a noise filter that’s up to 20× stronger than default. While we still recommend carrying out a pilot experiment to determine the most appropriate level for your particular data, we have found that a setting of ×2 drastically improves performance while still giving good coverage of your data.

Seeing the effect in typical high-resolution data

High resolution instruments tend to produce a lot of low intensity, close-to-background signals that usually have no associated MS/MS. After peak picking, the ion map can be a dense mass of blue outlines indicating the ion locations (figure 1).

Ions found using the default filter strength Figure 1. Peak detection on data set using the default import filter strength

Ions found after double the strength of the import filter Figure 2. Peak detection on the same data set using a filter strength of ×2

Increasing the filter strength to ×2 has the effect of filtering out this background, allowing you to complete your analysis with a minimal drop-off in the number of proteins identified (figure 2).

To compare processing performance, we took a subset of files from a larger dataset and analysed them using the default import filter strength and also at a modified import filter strength of ×2. The following table shows the differences between them, giving processing time, RAM used, and the number of ions found at peak picking:

Stage Default import filter Modified import filter at ×2
Alignment (including automated reference image selection) Time taken: 5mins 14s
Peak RAM usage: 3.62GB
Time taken: 3mins 31s
Peak RAM usage: 2.29GB
Peak picking (default settings) Time taken: 27mins 46s
Peak RAM usage: 7.02GB
Time taken: 10mins 22s
Peak RAM usage: 2.97GB
Ions found 78,700 28,117

So, as you can see, both processing time and memory usage can be significantly improved by use of a filter strength more suited to your data. And while the reduction in the number of ions found may look severe at first glance, this is actually a positive thing; those were ions whose signal was at a level that should be regarded as background. Provided you’ve not increased the filter strength too far, the number of proteins identified should be practically unchanged.

Try it on your own high-resolution data

If you’ve got a data set that your PC is struggling with, why not put this to the test yourself; you can find full instructions for the tool in our FAQs. In the meantime, we’ll be busy working on other ways to improve the performance of Progenesis to keep up with the demand for analysis of increasingly complex data sets. 🙂