A study using an n=1 experiment with single standards at 0V CID reported that 70% of detected ions were in-source fragments (ISFs). www.nature.com/articles/s42... This finding was extrapolated to suggest that ISFs affect all metabolomics experiments to this degree. A counterpoint. rdcu.be/ebFwc 1/n - ThreadSky

pieterdorrestein.bsky.social • 89 days ago

A study using an n=1 experiment with single standards at 0V CID reported that 70% of detected ions were in-source fragments (ISFs). https://www.nature.com/articles/s42255-024-01076-x This finding was extrapolated to suggest that ISFs affect all metabolomics experiments to this degree. A counterpoint. https://rdcu.be/ebFwc 1/n

Comments

pieterdorrestein.bsky.social•89 days ago

While we agree that ISFs are an important consideration, studies that explicitly investigate ISFs have found their prevalence to range between 2–36% of detected features, highlighting significant variability depending on experimental conditions.

pieterdorrestein.bsky.social•89 days ago

A key issue is the misrepresentation of the 70% finding and its amplification in media and scientific blogs. https://www.science.org/content/blog-post/phantom-metabolites and https://theanalyticalscientist.com/fields-applications/the-dark-metabolome-a-figment-of-our-fragmentation. I believe that such important and influential outlets should represent science better than they did here.

pieterdorrestein.bsky.social•89 days ago

This led to claims that all untargeted metabolomics data consists of 70% ISFs, that no new metabolites remain to be discovered, and that ISFs lead to widespread misannotations and suggesting that the “dark metabolome” is merely an artifact of fragmentation.

pieterdorrestein.bsky.social•89 days ago

These claims have had real consequences, including the rejection of papers and grants, particularly affecting junior researchers. Given these implications, it is essential to provide a balanced perspective. And our counterpoint is not alone - see (found 10% ISFs) https://www.biorxiv.org/content/10.1101/2025.02.04.636472v1

pieterdorrestein.bsky.social•89 days ago

Incorrect annotation arises primarily from untrained or inattentive analysis. Modern data processing workflows already account for ISF-related errors. The extrapolation from a single experiment to claim that the dark metabolome is largely artifactual is not scientifically justified.

pieterdorrestein.bsky.social•89 days ago

The rapid and continued discovery of new metabolites—even in well-studied organisms such as E. coli, mice, and humans—contradicts this argument.

angelsi.bsky.social•87 days ago

For me, Giera et al. have raised an interesting point about METLIN data. Are those standards analyzed under ESI conditions prone to ISF (a worst-case scenario, ok for robust library?), or is the molecular space covered by METLIN (i.e. molecules we can synthesize and isolate) inherently prone to ISF?

pieterdorrestein.bsky.social•87 days ago

Even though not verifiable -as data is not available to look at- I would imagine that robustness of the METLIN MS/MS library is OK as the precursors are isolated and then subjected to high CID energies-so even if introduced under high ISF conditions, the precursor ions are still fragmented to create

angelsi.bsky.social•87 days ago

Sure! Perhaps building libraries under high ISF conditions ensures they work well with high ISF data while staying robust for non-ISF data. As an analytical guy, I wonder how ISF affects MS2 and if certain libraries have ISF-related bias (most from standard concentration & LOD—Occam’s razor?)

pieterdorrestein.bsky.social•87 days ago

Some classes of molecules will be more prone to ISF with low energy input that exists at “0V”. See phosphodiesters such as ATP (also well known for phospho peptides in proteomics) or glycosylations. Many other molecules as well. Tuning instrument for soft ion transfer and isolation can help.

Comments

Posting Rules

Reply