A report from the Institute of Medicine has highlighted the on-going challenge of translational proteomics with this line, “transforming the great promise of these new [‘omics] technologies into clinical laboratory tests that can help patients directly has happened more slowly than anticipated.”
So, it’s good to be able to share a publication by one of our customers addressing this challenge. This paper provides a roadmap for how to go from generating a set of potential biomarkers by label-free, MS1 intensity-based, proteomics to a reliable pre-clinical toxicology screen based on SRM.
The work involved re-analysis of archived rat liver samples from the PredTox collaborative project. You can see the sample sources below and more background to the project, along with early discovery research data, is available in a previous blog post and a poster presented at ASMS 2011.
The InnoMed PredTox Consortium is a partnership between pharmaceutical companies, small-medium enterprises, and academic institutions in Europe. It’s aim is to take a combined ‘omics approach to study animal models of pharmaceutical toxicity in an effort to improve pre-clinical safety evaluation.
The final publication is open access, so I don’t need to summarise the whole paper. Instead I’m going to pick out some specific aspects that may be useful to anyone considering the same approach.
The discovery phase
In re-analysing the rat liver samples the group took a number of steps to maximise quantitative and qualitative information and generate reliable results from label-free LC-MS, including:
- Optimising run alignment by injecting a pool of all samples, which provides a reference run containing all features
- Duty cycles of the MS instrument were optimised to maximise quality of MS1 scan data for reliable quantitation
- Pooled samples were analysed with duty cycles optimised to generate MS/MS data for maximum peptide/protein coverage, which are merged with the existing quantitative data
- Targeted inclusion lists were run to further increase proteome coverage by selecting MS/MS for significantly changing features not identified by pooled sample analysis
- Filtering selected charge states and retention time sections of the LC-gradient to improve data analysis by removing less reliable features from the analysis
- Combining data from individual peptides into a final list of quantitated and identified proteins determined as significantly changing based on fold change >1.5 and ANOVA p-value <0.05
- Benchmarking technical variance of the analysis method using six repeat injections of a pooled sample, which showed CVs <20% for all quantified MS1 features.
These steps are supported, directly or in-directly, by Progenesis LC-MS and its approach to data analysis. To see a full set of features and how they help analyse quantitative label-free LC-MS proteomics experiments, you can download Progenesis LC-MS and try it with your own data or the tutorial data set included.
The validation phase
The list of significantly changing proteins from label-free LC-MS analysis, along with data from a previous transcriptomics study and literature searches, provided a panel of potential biomarkers for hepatoxicity. These were used to create a reliable SRM assay, helped in a large part by Skyline.
This section of the paper highlights some of the technical challenges you face moving from an untargeted, label-free proteomics approach to a targeted, multiplexed assay measuring 10’s-100’s of proteins. These include:
- Having appropriate MS/MS data for the proteins you want to include in your panel, in this case the library of MS/MS spectra was build up using the label-free analysis approach above
- Selecting peptides of appropriate length, amino acid composition and suitability for SRM
- Only modest correlation between proteins showing significant changes in response to treatment by proteomics compared to transcriptomics studies
This shows why starting with an untargeted discovery approach using proteomics is necessary, since only a small percentage of potential biomarkers will be suitable for the final targeted analysis approach. In this case 717 proteins identified by label-free LC-MS, plus others included form other studies, translated into a final set of 48 proteins (<10%) in the validated SRM approach.
Want to know more?
This publication shows what can be achieved by applying the right technologies in the right way to translate proteomics from biomarker discovery into a robust assay for pre-clinical or clinical application. It’s especially pleasing for us to see Progenesis LC-MS playing a key part. If you would like to speak to one of our product specialists about how we can help with ‘omics data analysis for your research, please contact us.