How Does COMPASS Work?
After launch, users will be able to upload mass spec (MZML) files to be analyzed by COMPASS. Users will also need to upload a sample information dataset (CSV) file that contains any number of sample parameters. Users can then create one or more workbooks to analyze the mass spec data in relation to any or all of the parameters in their sample information dataset.
If users have data on the molecular identity of specific features within your MZML files, they can upload this reference information to create a dictionary. In subsequent analysis, COMPASS software will indicate when any high information biomarkers are likely annotated molecules from users' dictionary.
Once users have uploaded their mass spec data, a sample information database, and a reference map of feature identities, users are now ready to rapidly explore and visualize their data. A video demonstration of how our tools can be deployed to analyze a series of mass spec files can be found here (coming soon).
By selecting some parameter from their sample information database as an ‘outcome,’ users can then immediately rank millions of datapoints by how well they correlate with that outcome (their ‘decision value’) (Figures 1, 2, 3). If our standard resolution setting is used to extract the MZML files, this step is accomplished in minutes. After selecting high information datapoints, m/z, chromatographic retention time, abundance, and decision value are visualized. Users can zoom in to regions of interest or select individual datapoints to see how their expression varies across the sample set (Figure 4).
Datapoints with physical or biological connections will have expression patterns that are highly correlated across all samples. Users can perform cluster analysis on high information datapoints to reveal their connections, assembling them into ontology groups with connected biological functions. These ontology groups can differentiate samples into novel subgroups, perhaps revealing novel biology.
Users may employ one of several machine learning algorithms to determine how well their high information datapoints can predict their outcome parameter (Figure 5), and use that machine learning to discover those individual datapoints with the greatest predictive power. Users can visualize datapoint differences across categories in the outcome parameter to understand their behavior and identify different datapoints with similar behavior (Figure 6).
Combine multiple individual workbooks to compare information content of datapoints identified in different analyses or different experiments. This can reveal shared and unique biology for different parameters of interest.
Download lists of mass spec datapoints that can be used to focus mass spec molecular identification (MS2) experiments, revealing the identify of molecular species connected to biological interests.
Don’t follow months of mass spec data collection with years of painful data analysis. Use Magellan Bioanalytics’ computational tools to easily complete this data analysis in days. Importantly, deep expertise in mass spec experimentation is not required.