Chromatography Forum

Hi all,

I'm analyzing a dataset with Amdis. It's a big set of unknown samples ran in the same conditions, so my strategy is to match compounds across samples mainly based on the RT.

I thus started by setting the analysis parameters to be rather permissive with regards to the MS match (50 min match factor) but rather strict RT (RT window of 0.2 min, strong level penalty and 70 max penalty).
The logic is that I don't expect large RT shifts between samples, so I want to identify any compound which runs at the same time and is structurally similar to my target compound.

The problem is that Amdis seems to penalize components for no reason. For example, I have a component running at 3.766 min which is structurally very similar (weighted match - 99 and 98) to two components in the library whose expected RTs are 3.763 and 4.128.

I would like Amdis to associate the component only with the first target (RT diff 0.003) and expected the latter to be heavily penalized for the 0.36 min difference. However, both got exactly the same RETENTION penalty of -20.

Screen capture:
https://drive.google.com/file/d/0B7Rtha ... sp=sharing

Is this a bug? Or am I missing something?

Thanks in advance!

Seems like the same problem in:
viewtopic.php?f=4&t=27777

Thanks.
I could add that I can often eliminate the problem by running the analysis, rerunning it under different "level" setting and then rerunning under the original setting.

I will have to see what's more time-effective - switching to RIs or running the samples manually and making sure that there're no wrong penalties.

The Amdis project is no longer active, right? Short of these small bugs it's a great program.

AMDIS continues to be updated.

I think the latest AMDIS comes out with the latest NIST CD releases for the general populous.

But even AMDIS in NIST14 had the RT issue when I checked it.

It doesn't inspire much confidence if you need to reprocess under a different level to fix a bug,

Do you have the multiple compounds box ticked?, as it may help un-ticking it.

kind regards
Alex

Thanks again!

I'm happy to hear that it's still being updated. Given the low web profile and the fact that the last version available for download is from 2012, I thought we wouldn't see updates.

Do you have and details regarding the release of the next update?

I do have the multiple compounds ticked. The main reason is that I too often encounter peaks which Amdis interprets as two overlapping components which nonetheless have almost identical m/z profile. I usually work with very inclusive deconvolution parameters but still get multiple components which I judge to represent the same compound. In this case, I want Amdis to associate both with the same target.

BTW regarding the "multiple component": do you have any idea how the algorithm does that? Is it according to RT (i.e. if a good match is found, a component with later RT cannot be assigned to the target even if it's a better match)? Or according to overall match?

No, I don't know release dates of the latest version, v2.72 was in NIST14


From the AMDIS help file.

e) If Multiple identifications per compound is clear, a particular target can only be assigned as the best hit for a single component.

However, the target can still be assigned as a non-best hit for another component, but only where its match factor for the other component is less than the match factor of the best hit target assigned to that other component.

If Multiple identifications per compound is selected, there are no such “best hit” restrictions on assigning a target to a component and so the hit list adjustment process described above does not take place.

Your specifications are quiet varied, have you though of running a batch process, and using the
created batch file, open it in Excel to manipulate the raw output, so you can apply your own penalties.

kind regards
Alex

Thanks again.
I haven't thought of applying my own penalties but it's not a bad idea.

No, I don't know release dates of the latest version, v2.72 was in NIST14

Thanks for all the help, Alex.
Do you know whether v2.72 is available to download? All I find is 2.71.