Process Only set calibration from multiple systems

I'm having calibration troubles with PO sets that are made from sample sets acquired on different HPLCs. For example, i run 2 sample sets:
Sample set A- reference and some samples on HPLC1
Sample set B- just some samples on HPLC2
HPLCs are the same make and model, both sample sets acquired with the same instrument method, same processing method opened to enter calculation data in "Amount" window.
So I make a Process only sample set combining Sample sets A and B, with the idea that i'll use calibration from Sample set A to calculate results for samples in Sample set B. Processing is done with background processing, and a result set is obtained.
As i review the results, everything seems fine at first calibration curve is fine, results are calculated correctly. But once i get to samples acquired in Sample set B, no results are reported, because no calibration curve is found for those injections. It's as if Empower "forgot" the calibration curve when processing data from HPLC2.
On a seperate case injections were in mixed order,
Reference (HPLC1)
Sample(HPLC1)
Sample(HPLC2)
Sample(HPLC1)
The same happens here, as i'm reviewing results of injections from HPLC1, everything is fine, but no calibration is used on samples from HPLC2. For samples from HPLC1 at the end, calibration curve returns and correct results are processed. Is this expected behaviour for Process only sample sets? I don't see any reason for this, other that some samples are acquired on another system, as all other parameters are the same.


P.S. I am aware that i can use calibration curve from one sample set on another, it's just that getting all results in the same Result set fits better with company policy (as much as it can, considering the situation)

Answers

  • Another thing i noticed, when reviewing these results, for injections acquired with HPLC1 in peak calibration curve it shows that Y-axis value is area, but for injections from HPLC2 it's "Response"


    again, all data processed with the same processing method. And anyways, "Response" isn't even something that can be set as Y-value in processing method. 
  • I would investigate using a custom field to translate the "Response" and "Area" into a new field. The new field would be used in the "Y-Axis" of each processing method. This would likely be the easiest solution.
  • I see what you mean, but the thing is it doesn't matter what i set as Y axis value, for injections from HPLC2 it will switch over to "Response", just as shown in screenshots above.

    I can't find any source that would confirm that this is intended functionality or not, just seems like a weird thing to happen. I'd appreciate if someone could try this and confirm that it's not just our setup.
  • How can HPLC1 and HPLC2 have the same instrument method? Would they not be unique instrument methods for 2 different systems? I would check a few things in your process only sample set to make sure the functions and labels are matching up. First, is the method set the same for the whole sample set? So whatever processing method is in this method set should be applied to all the standards and samples? For the samples acquired using HPLC2, check the processing method- is the Y value Area, and X value amount? Are the component names spelled exactly the same in this processing method as the one used to generate samples from HPLC1?

    Process only sample sets are processed in the order from top line to bottom line, so it shouldn't matter that some HPLC1 and HPLC2 samples are mixed, its not processed in acquisition order. Check you have a clear calibration line at the top of the sample set. I know these things seem simple, but its worth double checking the following, too:

    Are the labels for all the samples from HPLC1 and 2 included in the Quantitate line at the end of the sample set. Are all these samples of the Unknown type? Is the Calibrate function and label references correct for your standards at the top of the sample set? Is the processing set to Normal for Clear Calibration, Calibrate and quantitate? When you background process, are you using Acquisition Method Set, Selected Method Set or just a processing method? If Acquisition Method Set, Empower could be using and applying only the method set from standards/samples acquired using HPLC1 and thus ignoring HPLC2. And if only using a processing method, remember that Empower ignores any function or labels or bracketing instructions when processing a sample set using a simple processing method- this option is really only useful when processing a selected set of standards, followed by samples, in the Injections or Channels tab.

    I would advise using "Selected Method Set" when processing the sample set and pick the method set used for the process only sample set- make sure that the processing method contained in the method set in the process-only sample set is suitable and matches the names in your component editor exactly. And again, make sure this method set is the one populated for all samples and standards in the process only sample set. If that still doesn't work, check peak codes in the result set- a lot of these are useful when troubleshooting a lack of results. Also, double check that the channels used for the HPLC2 samples aren't locked and unable to be processed in your process only sample set.


  • >How can HPLC1 and HPLC2 have the same instrument method?
    We've got a bunch of identical hardware setups 
    >First, is the method set the same for the whole sample set?
    Yes, on both separate sample sets and the PO. Therefore, peak name, Y value and so on stays the same. Usually we acquire data with method sets (MS) that contain only instrument method (IM) and process using specified processing method (PM). By now i've also tried MS with PM and processing with acquisition MS, nothing changed.
    Here is how my test PO looks, samples below the red line are acquired on another system, all information in component editor is nice and uniform, a single peak name with calculation data for all injections.

    Processed the data using acquisition MS, the first part of results are as i'd expect: 

    I've got calibration, results are being calculated, all is well. But when i go to the next injection:

    the calibration is gone. The only change i can notice is the different system. If samples acquired on HPLC60 in this case, the calibration would re-appear.

    Channels are not locked, as for peak and result codes- where can i decipher them, i can't find any explanation for  SC__ codes

    This DOESN'T happen, if i stitch together two sample sets acquired on the same system.

    Am i really the only one who gets a result like this?
  • Do you get a message in the 'Processing' tab of the message center when you generate a result set? The codes can be interpreted by searching the Help function, Search tab and put inverted commas around "processing codes", this will generate a list of articles, pick the Interpreting processing codes article and see does it match up what you get in your result set.

    Does your main peak have a curve reference selected in the processing method of the HPLC2 sample set?. Peaks which have a curve reference wont generate their own calibration curve. I haven't run PO sample sets before, so I can only guess Empower cant distinguish between different systems, even though in theory this shouldn't be an issue. Also, the fact that you ran HPLC2 samples as just samples and not standards may be the cause- stand alone samples with no calibration curve associated with them may not be quantified when you mix it with other channels.

    As an aside, I didn't know you could have multiple instruments set up under the one instrument method. Anytime I tried to do that, I got a message saying instrument methods needed to be unique and if I wanted to overwrite a previous method, I couldn't because data was linked to that method. I have Full Audit Trail turned on though. Hope the codes can help you.

  • >Do you get a message in the 'Processing' tab of the message center
    nope, no messages generated

    >Interpreting processing codes article

    and an example of the codes i'm getting:

    Not sure, are these the right codes? There are no SCxx codes in the Help article.

    >Does your main peak have a curve reference selected
    PM contains only the one peak, so i can't even assign a curve reference to it

    >HPLC2 samples as just samples and not standards may be the cause- stand alone samples with no calibration curve associated with them may not be quantified when you mix it with other channels.
    They are just samples. Making the PO set should associate the calibration curve to them, and it does if both sample sets are acquired on the same system. Even when mixing channels from 2 systems, channel name, description, wavelenght, everything is the same, as they are acquired using the same IM.

    >I didn't know you could have multiple instruments set up under the one instrument method
    we have about 20 2695+2487 systems working around the clock in >100 projects, having unique IMs for each system would be insane (and that's not counting h-classes, shimadzus, GCs and so on). 
    Or perhaps you meant it like HPLC1 is, for example, a 2695+2487 and HPLC2 is some other hardware, like e2695+2489? Because that was not the case for my data.
  • The message you are getting about invalid curve, this is what Empower states:

    "Appears when the slope of the calibration curve changes sign or goes through 0 between V0 and Vt".

    This could be one of those situations, like when making certain custom fields, where Empower simply doesn't work the way you want it.

    As for the instrument methods, we use Empower 3 FR2, and we have 20 UPLC instruments and 10 HPLC instruments. Every system must have a unique name when creating the system and associating it with a Node when configuring the system initially. So UPLC1, UPLC2, HPLC3 etc etc, and there are normally 2 or 3 systems linked to one Node, or LACE. When creating an instrument method in Empower 3, the first screen that appears is a list of current systems and their nodes, so we pick the instrument for example UPLC1, then call the method UPLC1-INS-Method, same for the other systems etc.

    Otherwise how is Empower going to know which machine you want to run samples on? If you have multiple HPLCs, like we do, and I want to run 100 samples on HPLC3 and 25 samples on HPLC8. I need a unique method set in the sample set for each sample set- one that has the instrument method for HPLC3, the other for HPLC8. Unless I'm missing something in how you set up instruments, this is how we use and run all of our samples, by first naming the instrument method after the system name, and then associating a unique method set containing that instrument method, so we have something to call up when running on the different systems.



  • >Empower simply doesn't work the way you want it.
    Suspected that, just wanted some confirmation

    Regarding the IMs, we are set up similarily, some 3-4 systems (so 6-8 instruments) per LAC/E32 node. If i go to create an IM, i select a system and Emp3 gets system configuration which is just what hardware the chosen system has, like this: 


    from there i just save the IM as "IM_1" or whatever, and create a MS using this IM (usually we don't add PMs to our MSs, so the MS ends up as just "IM_1_MS"). And then i can run sample sets on any system that has that exact hardware configuration, using my "IM_1_MS". So one MS works for all 20 of our 2690+2487 systems within that project. 

    If HPLC3 and HPLC8 in your example are identical setups, well... let's say i don't envy you :D


  • HeatherLongden
    edited July 2021
    I think the solution is in the channel name/ which in turn has some reliance on System Name. Channel names have to match for calibration to be allowed.

    I expect that the Response vs Area challenge is a part of this too. I asked one of our experts to review this thread

  • Hello - You cannot use a calibration curve generated from one system to quantitate samples from another system because of the difference in the system name.  That was intentionally done because there are numerous differences between the 2 systems, ie tubing, column, mobile phase, detector lamp ages and so forth.  So from a scientific view point I would not recommend it, however, there is a work around.  You could manually create the calibration table for the samples based on the calibration table previously generated from system 1.  All of the calibration points will be flagged as manually entered but it will allow you to quantitate the samples.  I hope that helps.  Neil