Peak Area and Retention Time algorithm

[sana_sankar]

I am new to chromotography. Infact I am a student of computer science.
I need to develop algorithms for Peak Area calculation and Retention Time.

which are the standard algorithms available ?Can anyone point to links which describe the algorithms. After two days of search, I could not find any such algorithms.

Regards
BSI

[mbicking]

The methods used by chromatography data systems are usually proprietary, so very little information is available to the public. However, I have done similar programming for simple applications and I can give you some ideas.

To get the peak area you must know the starting point and ending point for the peak. These points can be found by examining the first and second derivatives of the signal, and finding the point where the derivative exceeds some "threshold" amount, which is arbitrarily set.

The retention time is usually defined as the maximum point on the signal, which is easy to search for. You could also estimate it by using derivatives, or doing some curve fitting (second order polynomial works well) for the points near the apex.

The area is usually calculated as a histogram summation from the integration baseline to each data point, remembering to account for the data acquisition rate, which is set in the acquisition parameters.

This should get you started.

[lmh]

but integration is a big area. Have a look at standard texts such as "Numerical Recipes in C" (or whatever other language you prefer) as this has a whole chapter on integration of curves estimated by measurements or calculations at various intervals.

Remember, too, that you _cannot_ assume all the points are equally spaced in all instruments. (Numerical recipes covers that too... nice book!)

And there are other approaches not based on integration, too, but they are wild and seriously whacky, and I've never seen them used in real life. They include curve-fitting, and even a multivariate statistical approach (create a correlation between all measurements along a chromatogram and the target analyte!!!).

[sana_sankar]

Thanks a lot, lmh mbicking. I could get a bare idea of the algorithms.

Then from the net, I could find some material on integration algorithms, by waters (Empower 2 software: ApexTrack Integration)
Though the algorithms are mentioned, there description is at a higher level, of my understanding.

Now I have few more questions.

1. Why we need to compute baseline and how do we implement baseline algorithm?
2. How does the sofware detect peak detection and peak integration events ?

[Don_Hilton]

Why compute baseline: The signal for a compund of interest is the rise and fall of the detector signal from the flat background. You need to know where the flat background is to be able to measure the height and area of the peak. And that background is not always flat. you may have responses for other compounds or a gradual shift in signal. You need to be able to sort this all out.

How does the software detect peak detection and peak integration events: This is difference that makes various integrators work differently. Some examine the data points in sequence and look for indications of slope change and sufficient variation from baseline. Others look for peak apexes and look for patterns by examining data points moving out from the apex.

Spend some time with a chromatgorapher and have him/her show you what chromatograms look like and how they determine peaks by examination.

[sana_sankar]

I had an interaction with a chemistry professor of my university.
He believes, integration algorithms vary for chromotograpies. Thus peak area calculation used for liquid chromotography cannot be used for gas chromotography.
Is he right?

But I haven't read anywhere that gas and liquid have different peak area and retention computation methods.

My area of work is gas chromotography.

[lmh]

Gas and liquid aren't impossibly different, so some software (for instance Thermo's Xcalibur) works for both, and so far as I know uses the same integration algorithms.

There are differences if the peak-shape becomes very different. For instance, Water Empower has two integration methods, one of which is more suited to very sharp peaks (I am guessing they started thinking about it in connection with UPLC).

Capillary electrophoresis is particularly demanding, producing extremely narrow spiky-topped peaks which don't respond well to some standard integration methods.

On a practical matter, some forms of chromatography are prone to negative peaks as well as positive (it depends mostly on the detector). This can have a big effect on how you choose to find the baseline.

[lmh]

a couple of extra thoughts:
(1) your professor might have meant that the parameters of the integration algorithm need to be changed to suit GC or LC. GC peaks are narrower than conventional LC, and most algorithms like to know a few things like typical peak width before they start.
(2) You might like to have a look at the xcms software package, which is open source, written in the statistical language R. R itself is freeware/open-source (Google for R and "Cran" to find it; why do people use ungoogleable names like "R"?).

xcms is a metabolomics package, but it includes at least two algorithms for peak finding. I don't know how sophisticated its integration is (probably not very). But it is very much a growing thing, cutting edge in the metabolomics world, and if you get good at algorithms, I believe the xcms team are always open to suggestions and contributions.

Good luck!

[aceto_81]

a couple of extra thoughts:
...
(2) You might like to have a look at the xcms software package, which is open source, written in the statistical language R. R itself is freeware/open-source (Google for R and "Cran" to find it; why do people use ungoogleable names like "R"?)...

You might google for "R-project", this will give you the desired results (or use www.r-seek.org)

Also there is an intensively used R-help list which is very usefull both for starters and experts.

Ace

[cdsman1001]

I had an interaction with a chemistry professor of my university.
He believes, integration algorithms vary for chromotograpies. Thus peak area calculation used for liquid chromotography cannot be used for gas chromotography.
Is he right?

But I haven't read anywhere that gas and liquid have different peak area and retention computation methods.

My area of work is gas chromotography.

Empower does not differentiate between gas, liquid or capillary electrophoresis when it comes to integration algorithms.

[mbicking]

I had an interaction with a chemistry professor of my university.
He believes, integration algorithms vary for chromotograpies. Thus peak area calculation used for liquid chromotography cannot be used for gas chromotography.
Is he right?

But I haven't read anywhere that gas and liquid have different peak area and retention computation methods.

My area of work is gas chromotography.

ChemStation also uses the same data analysis software. As noted, what he may mean is that the individual parameters (threshold, peak width, etc.) will be different.