RETENTION PROJECTION

Retention times, predicted — not measured.

Chromatogram ridge families across temperature programs, computed
Scroll to descend
Retention ridges across temperature programs
Detail of one migrating peak
The projection · k vs T

Every compound keeps its schedule

A series of temperature programs, back-calculated through effective temperature and hold-up-time profiles, recovers k-versus-T relationships — without isothermal tedium.

The detail · resolution

Under two-fold error, across instruments

Accuracy comparable to isothermal measurement, and an error distribution that stays predictable across laboratories, instruments, and methods.

Fig 2.1 — Projection · k vs T, computed ridges
01

The problem with shared retention data

Direct methods to measure k-vs-T relationships are tedious and time-consuming. Shared retention databases are unreliable and highly dependent on the specific GC–MS system, and physical standards simply do not exist for many compounds of interest.

GC–MSLC–MSMissing standards
02

The back-calculation method

A new methodology uses a series of temperature programs and algorithmic back-calculation of effective temperature and hold-up-time profiles. It is relatively fast, easy, and requires no additional equipment — and it enables automatic calculation of retention-time tolerance windows.

Back-calculationTolerance windows
03

Accuracy that travels between labs

Retention projection is 3-fold more accurate under matched conditions, and 4- to 165-fold more accurate across laboratories using different methods — with a predictable error distribution. In LC–MS, 2- to 22-fold improvement over linear retention indexing.

Next: a large database of k-vs-T relationships, and faster methods for measuring k-vs-Φ to make the methodology routine.

4–165× cross-labk vs ΦRetention database
Retention projection ridges
Anchor publication — A practical methodology to measure unbiased gas chromatographic retention factor vs. temperature relationships J. Chromatography A 1374 · 2014
← The spectrum Machine learning →
B. Peng, Ph.D. — AppCubic · 2026 Imagery : Wellcome Collection CC-BY · RCSB PDB CC0 · computed in-house Google Scholar ↗