A friend joined a group studying some cell behaviour. They had previously had a big result that they could stimulate this behaviour in defined, serum-free culture by adding a specific factor.
Friend was to work on characterising this effect, so his first job was to reproduce the result as a base case. He couldn't. The factor didn't stimulate the behaviour.
He asked around, comparing his execution of the protocol with that of the the postdoc who had done the original work.
The method involved growing a feeder layer of cells, in serum, then lysing them and washing the plate, leaving a serum-free layer of extracellular matrix behind, as a foundation for the serum-free cell culture (this is a pretty standard technique).
Turns out the previous postdoc's idea of washing a plate was a lot less thorough than my friend's. Couple of quick changes of PBS. So they were almost certainly leaving a lot of serum factors behind on the matrix. Their serum-free culture was nothing of the sort.
The supervisor insisted that the previous postdoc's work was fine, and that my friend just didn't have good technique. The supervisor had him repeat this work for months in an attempt to make it work. But he's a careful worker, so it never did.
This is the worst situation when the supervisor (professor) “sees no evil, hears no evil”.
In a similar situation a prior students work couldn’t be repeated and it was pretty clear the student made up the results. “Water under the bridge, let’s move on”. Of course the publication still counted for the prof.
This feels like automation would have great benefits for these types of things.
Instead of relying on people getting the right technique, you load in their program, dump chemicals into the right vials, then let it run and check the results
Well, a lot of these tasks are already automated (ie, shakers), but most bench workers have their own quirks on existing protocols. Most labs have their own 'dialect' of common mol bio techniques that 'work' in their particular setup. Perhaps the reagent from their particular supplier requires a longer incubation time, or the enzymes are wonky and you need to add more. Everybody I know does washing steps their own way - they say the "official" protocol is too long/cumerbersome/wasteful. More often than not, their own variant of the protocol is documented in their lab books, but not in the publications, where it cites the original protocol.
I imagine the postdoc would have a negative control of not adding the vector? Otherwise its hard to convince people the effect was coming from the vector.
Friend was to work on characterising this effect, so his first job was to reproduce the result as a base case. He couldn't. The factor didn't stimulate the behaviour.
He asked around, comparing his execution of the protocol with that of the the postdoc who had done the original work.
The method involved growing a feeder layer of cells, in serum, then lysing them and washing the plate, leaving a serum-free layer of extracellular matrix behind, as a foundation for the serum-free cell culture (this is a pretty standard technique).
Turns out the previous postdoc's idea of washing a plate was a lot less thorough than my friend's. Couple of quick changes of PBS. So they were almost certainly leaving a lot of serum factors behind on the matrix. Their serum-free culture was nothing of the sort.
The supervisor insisted that the previous postdoc's work was fine, and that my friend just didn't have good technique. The supervisor had him repeat this work for months in an attempt to make it work. But he's a careful worker, so it never did.