by
lmh » Thu Feb 15, 2024 2:04 pm
I'm convinced incompetence is more common than fraud. If you take a random paper with data derived from a chromatography method, the likely faults will be:
(1) paper fails to describe how the chromatography was carried out. It will tell you the exact system that was used, but fail to mention the column, or it will state the solvents but not the gradient, etc. etc.
(2) method doesn't actually have any retention; the author is very proud of having developed a "rapid method for screening X", without realising that no retention means no separation, which probably means wildly inaccurate results because of coeluting stuff.
(3) method run wasn't the same as the method written. This can happen in several scenarios. For example, PhD student who did the work got the method from post-doc A, but both student and post-doc have moved on. Professor asks post-doc B for the method, who provides the method they use, unaware that post-doc A had a different method. Or, method was originally developed by post-doc A, who wrote it up as a standard operating procedure and stuck it in the lab's methods folder. Since then, it's been updated and modified over many years, and has diverged from the standard method, but when the boss asked for the method, someone reached for the folder...
(4) method relies on an ancient, dirty column and doesn't work on a nice clean fresh one. This happens either through dirt-analyte interaction chromatography (DAIC), or, more innocently, where the analyte is naturally sticky and tends to bind to metal surfaces. Once the column has been used for a few weeks with dirty samples, all possible binding-sites are occupied, and the method works fine.
(5) peaks weren't peaks. The professor was determined to see something, and hasn't looked at a chromatogram since they were age 22. The lab doesn't generally do chromatography, but a reviewer said you can't claim X has a role in Y unless you've actually measured X, so under pressure of resubmission deadlines, someone rushed a quick analysis of X, which didn't actually find anything because the extraction went wrong. The professor made the post-doc look closer and closer at a bumpy baseline until they found something that might look like a bump, not too far from the expected retention time, so that must be right, mustn't it? Variant on this: a processing method labels the bump at 4.5min as 2,3 dihydroxy chickenwire based on its being found at 4.5min, and the student believes it must be 2,3 dihydroxy chickenwire, not realising that the label means nothing more than "thing that eluted within 0.1min of 4.5min".
(6) peaks weren't peaks, a rare version: professor hasn't looked at a chromatogram since age 22, but has told PhD student to do some chromatography. PhD student doesn't know what a peak should look like, and found a spike at 5.6 minutes, a single point in the chromatogram that's higher than everything else (an air-bubble got into a detector? Who knows!). Student is blissfully unaware that real peaks have a distinct width and consist of multiple measurements. If someone's truly creative, they've used a Gaussian smoothing function on the data and converted the one high reading into a nice Gaussian curve. Since the material is "precious" and "difficult to obtain", and the student was in their last 3 months, there was never any replication, so the entire paper hinges on one piece of noise.
(7) standard was the wrong concentration. Post-doc is not numerate, but professor doesn't know. Post doc doesn't know the difference between moles and molar. Or can't read previous post-doc's handwriting. The standard is so outrageously expensive that it was bought once, 15 years ago, and someone made up the whole 100ug in 1mL of methanol, which has been the lab's primary stock ever since. The methanol has been evaporating, the original 1mL was actually 900uL because the pipette dribbled, and the student who last did a dilution from it got their calculations wrong, but because no one else could be bothered to do another dilution, the entire paper is based on one sub-stock that isn't what it said it was.
(8) standard wasn't even the right chemical. The authors are probably innocent of this one. They specifically wanted to measure 2,3 dihydroxy chickenwire, rather than the more common 1,2 dihydroxy chickenwire, so they ordered the standard from CouldntCareLessChemicals.com who looked it up and found a cheap bulk supplier of 1,2 dihydroxy chickenwire, and supplied that instead. Authors don't have NMR and aren't chemists, so they've been working on the wrong stuff ever since.
There are lots of other things that can go wrong. One of my least favourites is naively optimistic belief in the reliability of library matches on spectra. My other least favourite is measuring until you get the expected result, and declaring that all other experiments "went wrong".
