Glossary
Signal to noise ratio
Signal to noise sets the floor of what an instrument can see in a single run. How you treat that floor decides whether real low abundance features are recovered or discarded.
Definition
The signal to noise ratio compares the intensity of a peak against the magnitude of the background fluctuation around it. It governs whether a feature is judged detectable, because peaks near the level of the noise are hard to distinguish from random variation in a single measurement. Thresholds are often expressed as a signal to noise cutoff, which makes the ratio a practical control over the sensitivity and specificity of detection. Because noise is random and does not reproduce, agreement across replicate injections provides evidence that a low ratio feature is real even when any single run is ambiguous.
Worked example
In practice
A peak just above the noise in one injection is uncertain, but the same peak appearing at the same m/z and time across ten injections is very unlikely to be random, which raises confidence without changing the instrument.
What matters
Where this makes a difference
A single run limit
In one measurement, low ratio features are genuinely ambiguous. That is why a per sample threshold is conservative by default and why it discards the weakest real signal.
Reproducibility beats thresholds
Noise does not repeat. Consistency across injections is stronger evidence than raw intensity, so cohort agreement recovers weak features a fixed cutoff would reject.
Relation to detection limits
Signal to noise underlies common definitions of the limit of detection, so how it is handled directly shapes what a study can and cannot find near the baseline.
Keep reading
Related
Want to see the method on real data?
Read what a dataset assessment involves and what you get back.
Grounded in the same first principles that Metablify is built on.