Metabolomics offers the opportunity to characterize complex diseases. The use of both LC-MS and GC-MS increases the coverage of the metabolome by taking advantage of their complementary features. Although numerous ions are detected using these platforms, only a small subset of the metabolites corresponding to these ions can be identified. The vast majority of them are either unknowns or “known-unknowns”. So we propose an innovative network-based approach to enhance our ability to determine the identities of significant ions detected by LC-MS. Specifically, it uses a probabilistic framework to determine the identities of known-unknowns by prioritizing their putative metabolite IDs. This will be accomplished by exploiting the inter-dependent relationships between metabolites in biological organisms based on knowledge from pathways/biochemical networks. This is the R package MetID that implements the algorithm.
The main function in this package is get_scores_for_LC_MS. See ?get_scores_for_LC_MC for documentation. This function takes an input dataset and assigns scores for each putative identifications. When working with this function, you must:
Have a data file with .csv or .txt extension. Otherwise, you need to read it in R as a ‘data.frame’ object first.
Check if the colnames of your data meet requirements: columns named exactly as ‘metid’ (IDs for peaks), ‘query_m.z’ (query mass of peaks), ‘exact_m.z’ (exact mass of putative IDs), ‘kegg_id’ (IDs of putative IDs from KEGG Database), ‘pubchem_cid’ (CIDs of putative IDs from PubChem Database).
This example shows the usage of function get_scores_for_LC_MS with a small dataset: demo1. This dataset only contains 3 compounds and is documented in ?demo1. Note: the scores are only meaningful when we have a dataset with a large number of compounds. So the result of demo1 dataset does not make sense.
data("demo1")
dim(demo1)
#> [1] 20 6
head(demo1)
#> Query.Mass Name Formula
#> 1 371.2283 sn-Glycerol 3-phosphate bis(cyclohexylammonium) C15H35N2O6P
#> 2 371.2283 1,4-Bis(chloromethyl)-2,5-diheptylbenzene C22H36Cl2
#> 3 450.3221 N-(3�_��,12�_��-dihydroxy-5�_�_-cholan-24-oyl)-glycine C26H43NO5
#> 4 450.3221 Glycodeoxycholate C26H43NO5
#> 5 450.3221 Deoxycholic acid glycine conjugate C26H43NO5
#> 6 450.3221 Chenodeoxycholic acid glycine conjugate C26H43NO5
#> Exact.Mass PubChem.CID KEGG.ID
#> 1 370.2233
#> 2 370.2194
#> 3 449.3141
#> 4 449.3141
#> 5 449.3141 22833539
#> 6 449.3141
Since the colnames do not meet our requirement, we need to change its colnames before we use get_scores_for_LC_MS function.
colnames(demo1) <- c('query_m.z','name','formula','exact_m.z','pubchem_cid','kegg_id')
out <- get_scores_for_LC_MS(demo1, type = 'data.frame', na='-', mode='POS')
#> Start building network: it may take several minutes......
#> Start getting random samples: it may take several minutes......
#> Start writing scores......
#> Completed!
head(out)
#> query_m.z name formula
#> 1 371.2283 sn-Glycerol 3-phosphate bis(cyclohexylammonium) C15H35N2O6P
#> 2 371.2283 1,4-Bis(chloromethyl)-2,5-diheptylbenzene C22H36Cl2
#> 3 450.3221 N-(3�_��,12�_��-dihydroxy-5�_�_-cholan-24-oyl)-glycine C26H43NO5
#> 4 450.3221 Glycodeoxycholate C26H43NO5
#> 5 450.3221 Deoxycholic acid glycine conjugate C26H43NO5
#> 6 450.3221 Chenodeoxycholic acid glycine conjugate C26H43NO5
#> exact_m.z pubchem_cid kegg_id inchikey metid score
#> 1 370.2233 <NA> 1 -
#> 2 370.2194 <NA> 1 -
#> 3 449.3141 <NA> 2 -
#> 4 449.3141 <NA> 2 -
#> 5 449.3141 22833539 WVULKSPCQVQLCU 2 0.500666666666667
#> 6 449.3141 <NA> 2 -
We also include a large dataset (demo2) which generates meaningful scores. As well as data frames, MetID works with data that is stored in other ways, like csv files and text files.