How RI-to-RT prediction works in GC-QQQ method building
RI-to-RT prediction is a practical calibration step, not a magic correction. The current workflow uses your own alkane RT data to map RI references to predicted RT values, then attaches RT windows and warnings so you can review the result before export or routine use.
Direct answer
RI-to-RT prediction works by taking a compound's RI reference value, calibrating that reference against measured n-alkane RT values from your own method, and converting the RI into a predicted RT that can be reviewed in the same workflow as the rest of the transition table.
In the current implementation, this is useful because RI is more portable than raw RT, but still not portable enough to skip calibration. The prediction step reduces manual RT entry, yet final RT suitability still depends on your instrument, column, oven program, and lab conditions.
Build transitions first
Start with matched compounds and generated rows that already carry RI references when the active workflow supports them.
Enter alkane RT data
Provide a monotonic set of C8 to C35 alkane RT values from your own GC setup to calibrate the prediction model.
Map RI to RT
The current implementation uses a monotone piecewise-linear calibration model to turn RI references into predicted RT values.
Review warnings and windows
Check out-of-range warnings, RI-missing fallbacks, and RT windows before treating the result as method-ready.
RI vs RT: practical difference
| Aspect | RI | RT |
|---|---|---|
| What it represents | A normalized retention reference stored in the library or method dataset | A measured retention time under your actual GC conditions |
| Where it is used | Library matching, prediction input, method-aware reference data | Final acquisition review, windows, and method suitability checks |
| Portability | More portable than raw RT across similar workflows | Directly tied to your instrument, column, program, and conditions |
| Main limitation | Still needs calibration against your actual method conditions | Cannot be borrowed blindly from another lab or another setup |
How the current calibration works
The current generator workflow accepts alkane RT input in the form C8, 2.466, one line per alkane, and validates that the data stays within the supported C8 to C35 range. At least 5 valid alkane lines are required before calibration can run.
Once the alkane input is accepted, the current implementation builds a piecewise_linear_monotone calibration model. Each compound RI is then interpolated between the surrounding alkane RI points to calculate an `RT_pred` value.
If a compound RI falls below or above the calibrated RI range, the current logic returns the nearest edge RT and attaches a warning. That is useful as a fallback, but it should be treated as lower confidence than an interpolation that stays inside the calibrated range.
What the output updates
- fills `RT_pred` for compounds with usable RI references
- assigns RT windows, using existing window text or the default fallback window
- returns warnings for out-of-range RI values
- can fall back to method RT when RI is missing and a method RT is available
Where calibration input comes from
The calibration input should come from measured alkane RT values under the same GC method context you care about. That means the same instrument conditions, column setup, and oven program that you want the RT prediction to reflect.
In practice, this is why RI-to-RT prediction is useful: it gives you a consistent way to project RT from library-scale references into your own method environment. But the quality of the result depends on the quality and relevance of the alkane calibration you provide.
Common failure points
Insufficient alkane points. The current workflow requires at least 5 valid C8 to C35 alkane RT lines.
Non-monotonic RT input. Alkane RT values must increase monotonically, or the calibration is not valid.
Out-of-range RI values. A prediction outside the calibrated RI span should be treated as a warning case, not a high-confidence RT.
Confusing fallback RT with calibrated RT. A method RT fallback is useful operationally, but it is not the same as a true RI-based interpolation.
Use the trust pages together
This guide explains the prediction step. For the active dataset scope, read the coverage page. For the normalization and build pipeline that surrounds prediction, read the methodology page.
If you are applying RT prediction inside a live workflow, the generator remains the main operational surface. If you already know your family, you can also start from the category pages for pesticides, environmental, or odor/VOCs.
Validation limits
RI-to-RT prediction reduces manual work, but it does not remove the need to confirm final RT behavior in the lab.
Users still need to validate RT windows, instrument suitability, matrix effects, and acquisition conditions before routine use.
FAQ
What is the practical difference between RI and RT?
RI is a normalized library reference tied to a compound and method context, while RT is the measured retention time on your actual GC method. RI is more portable than RT, but RT is what you finally review in the lab.
What alkane data does the current workflow expect?
The current generator workflow expects one alkane RT per line in the format "C8, 2.466" and accepts alkanes in the C8 to C35 range. At least 5 valid alkane points are required before calibration can run.
What happens if a compound RI is outside the calibration range?
The current interpolation logic clamps to the first or last calibrated alkane RT and returns a warning that the RI is outside the calibration range. That result should be treated carefully and reviewed in the lab.
What happens if RI is missing for a compound?
The current calibration API can fall back to a method RT when one is available, and it returns a warning that RI is missing and a method RT fallback was used.
Does RI-to-RT prediction replace laboratory validation?
No. RI-to-RT prediction reduces manual setup and gives a more consistent starting point, but users still need to verify RT behavior, RT windows, matrix effects, and acceptance criteria on their own instrument.
Ready to calibrate RT predictions?
Build your rows, load alkane RT data, run calibration, and review the updated RT predictions and windows before export.