disparity on views of experiments, details and interpretations of data

Since I started in science I've been trying to not have prejudice against certain types of people I meet and collaborate with. It's been working so and so. I have to admit that I have, quite often, said to myself "open mind sweetie, have an open mind maybe they do understand the idea with science and the need of details".

To make this a sweet and short post without pointing fingers, I have collected shorter key issues to explain some of my frustrations with specific instances that come up more often than I would like it to. Of course, I'm simplifying the questions and the answers but most of it is, imho, fairly general.

Do I draw conclusions from an experiment with an n=1?

If you do an ELISA, drug screening or similar in vitro screen with cells - you need at least one bioreplicate to have any idea that the data is really real. (Exception - if you have a validated GMP assay where all the parameters are locked down and specific to the conditions and cells/material you are testing. This type of validation isn't really what I see in research on an every day basis.) I personally look "is the assay valid?" (did the positive/negative control work, are the CV% working etc). Then I would wait for the bioreplicate to state "this is the result" to a larger audience and make my decisions of future research on it. And, it would be recommended to have a larger n within each run, like having triplicate wells/plates per run. In the end you have two (or more) data points from triplicates in bioreplicates, which gives you better stats and more importantly, will show you more the variability within the assays.

How do you define "one experiment - one data collection"?

If you find yourself using a complex machine (with exchangeable parts that differ every so slightly between themselves) - you shouldn't consider it the 'same experiment' if you exchange one of these parts in the middle of the run of the machine. Say for example, changing pipettes in the middle of an ELISA run or change the column in an HPLC run, or switch a pin tool in the middle of a run. For analysis reasons and data variation, you should minimize the various parameters between the experiments and keep them consistent through out the run.

How much details in written protocols involving computer programs used for collecting data and interpreting said data do you need?

Short answer: everything needed in order to replicate the experiment exactly.

Long answer: it sucks, I know. You have to write down the settings on the plate reader (double reads, single reads, wave length, columns first, row first), gating of the cells for the flow analysis, number of dips with the pin tool, number of dilutions ("in and out of the pipette") etc. A lot of this can be solved from doing the same procedure every time. A number of people I know have a generic system that they keep with every time, "I always aspirate/disperse with my pipettes 8 times when I do the ELISA". And when it comes to plate reading, the recommended course of action is to save a protocol with the settings and use the same protocol every time for the assay read. However, to not document what the settings is... let's say not recommended.

Last pièce de frustration, if you are embarking into a - for you - unknown territory and work with someone whose area of expertise is said territory..... I would highly recommend you ask questions and listen to the answers to learn new things. One of the best traits in a scientist is humility and being humble about all these details that we don't understand nor know. And that it takes team work to make progress in complex questions. Part of that team work? Listening to other people who are experts in their subfield.

There. I'm done on my soap box now. Off I go to another day in the life of a scientist :)