Scenario: You have an advanced degree in STEM and as you sign off for the night, your boss texts you to put together a journal club tomorrow morning on That Thing You Mentioned. Unfortunately, That Thing was something you had given a critical caveat: You’re no expert in this, but your friend knows the folks who came up with it, so maybe…! Alas, the task has now fallen to you. It’s nearly time for dinner, there are chores, and you have to hit the gym, so you have maybe ten hours before the morning meeting to slap this thing together and at some point in there, you should probably sleep. For the sake of argument, your friend who knows That Thing is not available. Also for the sake of argument, the assignment coming via text after 5 pm is extremely cool and good and normal. You want to get this done. Therefore, you have to become an expert in something you vaguely understand and be able to present competently on it.

Let’s get to work.

Order of Operations

Most broad disciplines, but especially biology, exist in their present states after myriad fractals of evolution, ideation, and invention arising from base states, to the point that finding those base states is only possible if you spend, say, 4-12 years at a university studying the past several eras’ developments. It is as though you gaze upon the canopy of a rainforest from above and someone has cleverly stuck out flags titled “Figure 1,” “Figure 2,” “Figure …” and “Figure n” to mark the path of their research, and you’re about to tell all your coworkers what those flags are and what they mean.

First of all, give kudos to those researchers for their canopy-flagging skills. Once congratulations have been sent, though, bring your focus down from the canopy. You need to understand the researchers’ perspective because you don’t have 4-12 years of knowledge and intuition connecting these Figures for you. Remember, you’re not an expert in this. You’re an expert in learning.

You might already know not to read primary scientific research articles from top to bottom like a novel. My recommendation for prioritizing foci to understand a paper shifts dramatically if I am not the local expert.

Why the difference? If I already understand the state of research in a topic, the title efficiently tells me 80% of the paper. A well-written title for a primary research article is the most important, best-supported finding in the article and is the summary of several years of work. I’ll check the authors and affiliations for names I recognize and I’ll scan the abstract for the hook that makes me interested in the paper. I skip the introduction or briefly skim it because I already know the topic. Then, I can jump into the meat of the paper, whine about figure design and error bars, and quickly extract the strong and weak conclusions. Lastly, I’ll skim the discussion to see if I misunderstood something in my haste, and bam, I’m ready to bang out a dozen slides for a journal article.

However…

First Principles – Authors, Abstract, Materials+Methods

We need to build up from below in order to be conversant in this topic and not just regurgitate jargon. If this field is unfamiliar, the title and abstract are full of vocabulary that isn’t (yet) useful for understanding the paper. Hunt down the professional profiles for the first and senior authors. What are their foci? What are the primary fields of knowledge used as the root structure holding up this research? Choose one or two, write them down, and pin them to the screen.

Fantastic! We now have the same viewing angle as the authors, even though they have a collective 20-80 years’ more experience. Now check the title and abstract. We want to extract the main hypothesis, the conclusion, and any relevant terms and acronyms. Deep knowledge of every term would take too long to collect, but with just a couple notes, the text becomes suddenly legible. Make a list, search the web (a few favorite sources are Gene Cards, Cleveland Clinic, and NCBI), jot down relevant definitions, and pin that list to your screen.

We now have root structures and a shared language. The authors said “the path turned left,” and we know what “the path,” “turned,” and “left” all mean. We need the toolbox, AKA the Materials and Methods section (which everyone should call the M&Ms). This is where we can finally flex our years of practical experience because half of these are probably familiar and most of the rest have at least come up in conversation. Scan the familiar ones for new terms – especially terms on the list now pinned on the screen – and move on. Make a list of methods that are new or need a quick refresher and slap it up there under the list of terms.

Lists in hand, we could check the references and research those M&Ms and that would be 100% fine and correct, but it might take a while to dig through yet more research papers that might not even have been cited correctly. This is where ye olde epistemological hygiene heuristics break down a bit because a good, rigorous STEM scientist does not need the same heuristics used to keep college kids in line: use Wikipedia and LLMs. People get tied up in knots about using these, but one place they shine is high-level explanations of established fields of knowledge. Practice due diligence obviously and maybe dig elsewhere for details (which, as a reminder, are not a priority at the moment). If a knowledge aggregator coughs up something that seems whacky, trust your instincts and go find an actual source. Use the skepticism born from experience to keep these tools effective and efficient.

Here is a sample LLM prompt that has worked well for me: “You are an expert in [field]. Succinctly explain [M&Ms] to a lay audience. Cite each piece of key information.”¹ Always check that those citations exist, of course, but this approach has been exceptionally useful for getting the big picture out of context of the paper in question. Plus, by verifying the citations, I can track down originating publications and, therefore, illustrative figures that might help in the opening slides to a journal club presentation.

I estimate that at this point, it has been less than half an hour – maybe as little as ten minutes – and you have a quick sketch of notes that looks something like this one I recently made for a writing test for a job application:

• Primary fields of knowledge
  • cancer immunology, esp. killer T cell evasion, esp HNSCC
  • microenvironment cell-cell signaling
• Key terms
  • TCA: tricarboxylic acid (cell metabolism)
  • HNSCC: head+neck squamous cell carcinoma
    • mucosal membranes; HPV+/-; cetuximab; often can’t cytotoxic chemo
  • CPI-613: small molecule TCA inhibitor (Phase III trial failure? that’s a spicy COI…)
  • thrombospondin-2: secreted protein, short-range
  • AKT-mTOR: signal transduction pathway, increase metabolism, proliferation, etc (mechanistic target of rapamycin (rapamycin used in paper?))
• M&Ms
  • Rag1-KO mice
    • no mature T cells/B cells
    • “non-leaky” immune deficiency
    • mostly healthy when kept right; big tumors when tumor’d
  • Cell lines
    • MOC2: murine HNSCC model
    • HN6+HN12: human HNSCC
  • TF disruption: dead Cas9 w/gRNA vs consensus binding site + flanking blocks TF binding
  • IF quantification: avg 10 fields of view x 3 investigators (nice!)

NOW we can read the paper. We have the authors’ focus. We have their vocabulary. We have their toolbox. Instead of hopping from bit of jargon to acronym to esoteric graph and juggling alien terms, we can look up from the tree trunks at those cleverly tied flags and integrate years of experience into this new web of knowledge.

And! You still have like nine and a half hours before you need to give this journal club. I’ll bet you could even squeeze in a game of League of Legends or an episode of Bones.

What Were They Thinking? – Title, Introduction, Discussion

“Ugh why would they ever draw the graph that way? That’s a terrible WISH stain! Did they just not have a statistician on sta-”

Shhhhhh shh shh shh why are you looking at the figures and results? The authors are still better at this than you. This is their field – there are still six hours until this has to be your field. Let’s get behind the eyes of the authors and find out what they think was cool about the research.

The utility of the title is perhaps obvious, and now that we understand the lingo, we have a pretty good idea what it means. Every sentence of the paper should contribute ultimately to supporting the title. If the authors were good, they stayed on topic, and so should we. To continue our rainforest metaphor, the path started with the authors’ backgrounds and publication history, and it ends at the title.

In a perfect world, the introduction and discussion provide the classic hourglass structure to a paper: Start broad to narrow, end narrow to broad. That structure was especially available when research was more siloed and there was less ability to examine many-dimensional, interacting networks of variables because we barely knew what the variables were to start with. The trees holding up this canopy were treated as singular, so a paper could declare it had looked in the forest and found this tree with this branch and this leaf. It wasn’t easier or harder or better or worse; it was a necessary step on the path to comprehension.

In the modern era, we know that nothing is simple and nothing exists in isolation or as convenient, hard-lined categories. If we weren’t confined to two spatial dimensions and one visual field, it would be much, much easier to introduce a conflux of topics by simultaneously funneling information from three to seven fields into the singular synthesis the authors intend to achieve.

Until we can all read from n-dimensional, cone-shaped manuscripts, though, we can use this restriction to decipher the authors’ priorities and use the same lens they did to view the project.

In that last example, even though the research brings together multiple fields, the authors clearly frame the paper as one investigating genetic mouse models, with a few special focuses. Rare disease research challenges, skeletal biology, and CRISPR are all part of the problem and part of the solution, but those are secondary to the main goal of generating high-fidelity mouse models of human diseases. What does this mean for you, practically, in the remaining few hours before journal club? It means that when we finally approach the results and figures, it’s okay for rare diseases to simply be a class of diseases, for bones to be biological tissues, and for CRISPR to be a black box of molecular magic. Instead, focus your time and effort on how each figure contributes to the primary aim.

I cannot recommend spending as much time on the discussion section as on the introduction. Introductions are necessarily clean-cut guides into the research at hand; discussions are somewhat more laissez-faire. Discussions can speculate on contributions to other fields, they can pose possible alternative hypotheses, and they can opine, all of which are luxuries afforded to those writing full-length papers or opinion pieces, not those boning up on a subject as quickly as possible. The first few subheadings concisely state the benefit of the paper in context of the results, clue you into critical caveats that inform their downstream utility, and spell out what’s next. They reference specific data points of interest, so highlight those, because at long last, the results loom near.

You have achieved a tremendous feat: You started by staring at an arboreal constellation of seemingly unrelated words and cryptic graphs. You know now what the goals and impact of the research are. And it’s been, what, another half hour to hour? You probably deserve a snack and a stretch.

Grasp the Narrative – Figures and Results

This is the easiest part. As you come back from your snack and obligatory social media binge, you check the notes pinned to the monitor to recalibrate, then look back up at those soaring, metaphorical branches of fact and discovery and design. There is so much more to learn, but the purpose here and now is not to know every leaf and stem – it’s to have enough knowledge to guide a curious naif like you without getting lost.

The results and figures of most papers follow a format that was drilled into me time and time again by my mentors: A simple story. There are a beginning, a middle, and an end.

How to distribute focused time across these comes down to a judgment call. Maybe the validation experiments are rote and perfunctory, but maybe they are the crux of the paper and show off a new method; maybe the mechanistic experiments are color-by-numbers, but maybe they are elegantly designed and exciting; maybe the real-world application is niche and inconclusive, but maybe it really is a triumph. Spend a little extra time on findings that are personally exciting, and try not to get lost in the weeds of over-complex figures that will eat up too much brainpower.

Regardless of what we choose to emphasize, we still follow the same narrative flow the paper follows. There are a beginning, a middle, and an end. Each figure drives from the starting point declared by the authors and abstract towards the goalposts established by the title, introduction, and discussion. Take each figure title and rephrase it, then hold those words up against the text of the results and make sure they align. Those are the bullet points or entire sentences that will serve as guide posts for generating the slide deck or summary article or whatever final product is being made. If appropriate, include some critique or editorializing, but avoid the temptation to be dismissive or over-critical – you chose this paper, after all!

And how long does it really take to look at a few figures, check the figure legend, and translate that into a couple phrases or sentences? Fifteen minutes? Thirty? By my reckoning, it’s been an hour, and you now understand this paper well enough to educate any lay audience and a lot of specialists. Not only that, but you can write and speak with confidence. You are less likely to fumble the terminology or misidentify the novel impact, because you learned from the ground up and have more reference points than a couple disconnected branches. You even have all those notes I told you to pin to your screen.²

In the end, this is no different from how you would approach any other project.

1. Determine perspective and intention: authors, abstract
2. Array your tools: definitions, M&Ms
3. Mark the path: title, introduction, discussion, results
4. Go.

Before I Head Out…

Flex your skills.

Build from the ground up. Speak from sure footing. It will show in your writing and in your voice and in your body language. Learn the words, learn the concepts, shirk the frills, tell a story.

Some papers aren’t great. Your job is to know them well enough to know their strengths and their flaws. It is not your job to fix them or make excuses for them.

Writing is hard. Learning is hard. You will mess up and be wrong. But! Do not lose the name of action. You are an expert. Just not at that Thing (yet).

Will Towler

I do… SCIENCE? and games…? and write. and throw people sometimes.

Thanks to The Quilled Sister for critique and editing. She also writes things!

1. This has some flaws. Often, if you list out the materials and methods in your paper of interest, the LLM will find that paper and cater so directly to it that the definitions lose utility. You may want to prompt each one individually, using an “edit prompt” function or iterative “ignore previous prompt” queries. Also, be diligent. You must not do this if what you’re investigating is under embargo or is otherwise protected intellectual property. And never assume it “knows” things. Repeating what I said in the main text: If an LLM (or Wikipedia) says something suspicious, trust your instincts and abandon the LLM for that topic. ︎
2. This practice is not good for your screen. I cannot actually recommend it. Maybe Sharpie? ︎