Hacker News new | past | comments | ask | show | jobs | submit login
Laboratory Notebook Skills [pdf] (dur.ac.uk)
91 points by Tomte 26 days ago | hide | past | favorite | 20 comments

Enabling reproducibility is a costly investment, but the cost of doing unreproducible work is even higher. This is a calculation I rarely see organization make.

More frequently everything is a one-off, with failure recurring randomly until a working (though not reproducible) approach is found. I'm not sure how to spin this evolutionary approach positively, as it can result in fragile outcomes.

There is software infrastructure that enables reproducibility (virtual machines, and more recently containerization), but this concept rarely makes it to written documentation.

> Enabling reproducibility is a costly investment, but the cost of doing unreproducible work is even higher. This is a calculation I rarely see organization make.

Hopefully things like Alex Freeman’s http://science-octopus.org (or something like it) will take off very soon, which addresses this well (and will make science more accountable).

Meet Octopus, a new vision for scientific publishing: https://www.science.org/careers/2018/11/meet-octopus-new-vis...

Dr. Alexandra Freeman | Octopus: a radical new approach to scientific publishing - 29 October 2018: https://www.youtube.com/watch?v=Af6aITLEoD8

Thanks for the links. Sounds like an audacious effort.

Pivots away from the current approach are very difficult. Even minor deviations like https://arxiv.org/ take decades to gain "normal" status in the science community.

I did my Physics undergraduate at Durham, and was given this checklist as a first year. It is so surreal seeing it again on HN.

I really enjoyed having a lab-book to organise all my notes through my degree. It took time to write notes our properly, but it quickly paid off when having to check calculations, assumptions and writing up experiements.

That's awesome, love HN has relevant people around.

Could you shed some light on what the 'script' is in 'The lab book is NOT a copy of the experimental script'? Seems to reference homework like tasks 'Task 1: 100 ± 1 kΩ' but in a more undefined experimental context what would that be...?

Maybe "worksheet" is a reasonable equivalent. Describes the the needed context (e.g. what's the goal, what are the principles involved), the experiment itself (setup, steps to do, what to measure, ...), ... for a predefined experiment, especially in a classroom setting.

In the lab notebook you need the details what you did and what you calculated, but don't need to copy all the context for the theory etc.

Every so often I'll see something like this on HN that suggests that software engineers should be keeping similar records of everything they do, if for no other reason than to have a record of problems they've solved that can be grepped when a similar problem shows up. This seems like a good idea.

I wish I could be bothered to actually do it though

Every time you look up something on StackOverflow, refer to the API docs, or refer back to the ticket, use case, or requirements document, make a note of your question and the answer. Even when you just stop typing to take a break, note down where you are and what you just did.

Every time you think you need a TODO comment, make a note instead, or also.

Consider Kent’s Beck’s recommendation to write down every decision you make.

Some other views



I know all of this, but it's still a lot of effort that I'm apparently not willing to go to.

However I do at least partially do this. Whenever I use SO, I put a link to the answer in a comment next to the code that implemented it. Whenever a ticket or API doc requires doing weird stuff, I put a link to that too. Whenever I make decisions that might be slightly weird, I put a couple of paragraphs in the git commit.

Not the same, but still better than nothing I guess

> it's still a lot of effort that I'm apparently not willing to go to.

Just like any habit, you have to start doing it, build up, and make it so you can't not do it. It's not about discipline, it's about routine.

As an ops person/SRE, I can say that pretty much every person I know in this space ends up with these sorts of journals of things they’ve seen go wrong. The really useful, and harder, part is getting everyone to put them in a public area so you can go off of other people’s experiences as well. Having everyone’s barely-documented garbled mess of previous experiences can sometimes be a life saver.

Having an easily-searchable version-controlled record of my "things I'll have to Google again" has been incredibly useful for me.

The current incarnation of my "lab notebook" is here, and as of today 1002 days old: https://serhii.net/dtb

The most surprising thing for me was it's ROI was much higher than any other bookmarking / PKM stuff I tried, and I tried a lot. I had/have a semantic mediawiki for links, with tags, categories, estimates of complexity and ratings etc etc etc, but apparently searching through a long textfile or browser page is still much faster

Somewhat of a tangent, but:

> Top wave: power source, bottom wave: through resistor

Am I going crazy or are these waves at different timescales? I've never seen a scope display information like this before. Generally, I've seen multiple source voltage scale, but the time scale is the same for each input.

Or am I forgetting something...?

The time scales are the same. Top wave is voltage. Bottom wave is power. Power is twice the frequency of voltage.

Ha, that makes a lot more sense. Thanks, pwr-electronics :D

Wondering what a modern, digital version would be? Imagine a repo could work for the time-stamping and not deleting mistakes (immutable record). Images, annotations (i.e. 'Circle rogue points') seem much harder.

Why does everything that works need a digital version to make it "modern"? The physical lab notebook solves the problem at hand.

"In late 1987, Robert Malone performed a landmark experiment. He mixed strands of messenger RNA with droplets of fat, to create a kind of molecular stew. Human cells bathed in this genetic gumbo absorbed the mRNA, and began producing proteins from it.

Realizing that this discovery might have far-reaching potential in medicine, Malone, a graduate student at the Salk Institute for Biological Studies in La Jolla, California, later jotted down some notes, which he signed and dated. If cells could create proteins from mRNA delivered into them, he wrote on 11 January 1988, it might be possible to “treat RNA as a drug”. Another member of the Salk lab signed the notes, too, for posterity. Later that year, Malone’s experiments showed that frog embryos absorbed such mRNA. It was the first time anyone had used fatty droplets to ease mRNA’s passage into a living organism."


There are a handful of competent-to-good digital versions. A lot of accelerator-type facilities use PSI's ELOG: https://elog.psi.ch/elog/

I've become a big fan of the Center for Open Science's Open Science Framework: https://osf.io/

It isn't really a thing in my field, but it enables pre-registered studies, and it has some really lovely integrations with github, cloud storage, etc.

In my view, an issue for modern science is that much of what we do is already very close to being a narrative of itself, such as program code and the generation of data and meta-data by computerized instruments. Keeping a separate narrative will always require time and discipline, and can suffer from lack of upkeep.

I know scientists who still use paper notebooks. They are quite common in chemistry and other "wet lab" sciences, where it's actually hard to narrate at the keyboard while also wearing rubber gloves and writing with one hand while holding something with the other. I'm sure there's also a mental discipline about it, that might help them think.

When I was a science student, my notebooks were a disaster, and I paid the price by having to repeat work when I couldn't figure out what the hell I did. Writing was always awkward and painful for me.

Today, I use Jupyter notebooks. They are not a strict timeline type of narrative, but have the benefit that I can see if they are complete before taking a break by doing a "restart kernel and run all cells." And they are suited to the kind of work I do, lots of data and computation. And they are at least crudely searchable.

For my work I don't need time-stamping or an immutable record. I need to be able to figure out what I did, hours or years later.

There's a bunch of Electronic Lab Notebook platforms out there, someone already replied to you mentioning OSF. Others include Benchling and LabArchives. This Nature article is a good starting point, it has links to several universities' ELN guides. https://www.nature.com/articles/d41586-018-05895-3

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact