Reading scientific papers

For the discussion of the sciences. Physics problems, chemistry equations, biology weirdness, it all goes here.

Moderators: gmalivuk, Moderators General, Prelates

User avatar
Angua
Don't call her Delphine.
Posts: 5934
Joined: Tue Sep 16, 2008 12:42 pm UTC
Location: UK/[St. Kitts and] Nevis Occasionally, I migrate to the US for a bit

Reading scientific papers

Postby Angua » Mon Sep 19, 2011 1:21 pm UTC

So, I was wondering what sort of strategies people generally have when reading papers. I find myself getting bogged down in a lot of stuff that I don't understand, and having a hard time telling whether or not it is valid. I realise that a lot of this will come with future practice and knowledge about the subject at hand, but I'm supposed to read some papers of my choosing before I go back uni (I've chosen to do papers to do with immunity against helminth parasites, so that way they are all interconnected) and I'll be asked to answer questions about them in an exam (which thankfully doesn't carry any weight, except for letting my subject tutor get a feel for me, so I want to make a good first impression) but I don't know what I'll be asked.

So, what are the key bits you tend to look at - do you rank figures by their usefulness? How do you identify if the title accurately represents the contents of the paper? How useful are the introduction and discussion in trying to understand the results?

Anything else you find useful to consider?

edit: Also, what sort of questions do you think will be asked, considering that my tutor doesn't know what papers I'm choosing to read, except for the fact that they come under the umbrella of Infection and Immunity.

Also, if people want to use this thread to talk about papers from other subjects (eg I figure a physics paper would be vastly different) then that is also fine.
Crabtree's bludgeon: “no set of mutually inconsistent observations can exist for which some human intellect cannot conceive a coherent explanation, however complicated”
GNU Terry Pratchett

User avatar
D.B.
Posts: 197
Joined: Wed Sep 09, 2009 3:08 pm UTC

Re: Reading scientific papers

Postby D.B. » Mon Sep 19, 2011 1:53 pm UTC

How I read engineering-based papers (may or may not be useful):

I'd tend to use introduction/conclusion as a quick test as to how useful I think a paper's going to be, and from there decide how much time I want to spend reading it. As such they're generally the first two parts I read.

If I'm actually interested in properly understanding a paper, rather than just getting the gist of it and a rough handle on the conclusions, I find it very helpful to print the thing out rather than work with it on a computer screen. I wish I knew why that was - after all I'm looking at the same stuff. But with a printout I seem to read more slowly, and pick out the awkward little details more easily. I suppose I'd also make notes in the margin when doing this, but that doesn't explain the entire difference.

If it's a paper that I really really want to understand (e.g. I'm going to implement or present it) then writing a condensed 1-page summary of it as I'm going along really is the only way forward. The idea is to strip out the padding and get it down to the bare bones of useful stuff, make sure any maths/stats are written in a form I actually comprehend, look up methods that they reference but don't explain, etc. It might take a morning or so but usually by the time I've done this, unless waaaay out of my depth, I'll be at the stage I can talk sensibly about 90%+ of the stuff in the paper.


I've no real idea what kind of things they might ask you about papers they haven't chosen themselves. Seems like it would have to be very broad brush questions.

User avatar
Izawwlgood
WINNING
Posts: 18686
Joined: Mon Nov 19, 2007 3:55 pm UTC
Location: There may be lovelier lovelies...

Re: Reading scientific papers

Postby Izawwlgood » Mon Sep 19, 2011 2:02 pm UTC

What's the context here? Are you being quizzed on a paper, or are you presenting it, or are you reading it to increase your understanding of a field?

Personally, I don't yet have the comprehension level to read most Cell/Nature/JCB papers in less than an hour or so, so it's typically pretty slow going for me. I usually start with the abstract and the first half of the introduction to get a feel for what the paper is about, and then look through each figure to get a sense of how they support their claims and what directions they went with it (protein interacts, deactivated point mutant doesn't do this, point mutant mimics drug treatment, live imaging, some dynamics, etc).

The lab I joined has weekly meetings, and four of us alternative presenting, so this last summer and during my rotation, I presented a bunch of papers. For presentation purposes, I typically copypasta the figures into a Power Point, include notes to explain the figures, and use introduction and conclusion slides to present the model, give an outline of what is being done, explain any novel techniques utilized, and discuss which results I felt weren't very strong or were misinterpreted.

EDIT: As for dealing with details; the only time I really read the methods/materials section is if they are doing a technique I'm unfamiliar with. It is entirely irrelevant to a general reader what concentration buffer was used, which primers amplified a sequence, etc.
... with gigantic melancholies and gigantic mirth, to tread the jeweled thrones of the Earth under his sandalled feet.

User avatar
Angua
Don't call her Delphine.
Posts: 5934
Joined: Tue Sep 16, 2008 12:42 pm UTC
Location: UK/[St. Kitts and] Nevis Occasionally, I migrate to the US for a bit

Re: Reading scientific papers

Postby Angua » Mon Sep 19, 2011 2:10 pm UTC

We are supposed to spend this year learning how to do a more evidenced based approach, ideally so we can figure out what is good with new medicine in the future (as opposed to my last 2 years which was basically just being taught lots of science that has to do with the body and disease). It will be a written exam, 2-3 hours long, and half of it will be us being given a paper (with the title and abstract scrubbed out) and being asked questions about it, like what is this about, what would be a good title, what is the crucial figure, write a 200 word abstract. The other bit is going to be about the papers we read, which we got to pick, and I won't have them with me, so I'm a bit worried about what he expects me to know or talk about. I have always had a problem with summaries, because I have a hard time filtering out what is relevant, and what is just added fluff, because a lot of times I think some detail is important, when it isn't.

I have printed out the papers - I work better on paper too, and making notes.
Crabtree's bludgeon: “no set of mutually inconsistent observations can exist for which some human intellect cannot conceive a coherent explanation, however complicated”
GNU Terry Pratchett

User avatar
Izawwlgood
WINNING
Posts: 18686
Joined: Mon Nov 19, 2007 3:55 pm UTC
Location: There may be lovelier lovelies...

Re: Reading scientific papers

Postby Izawwlgood » Mon Sep 19, 2011 3:19 pm UTC

Personally, I'd say the hardest part about the exercise of writing an abstract for a paper is not simply parroting back the abstract you already read.

So have you not read many papers, has most of your knowledge accumulation been from texts or lectures? Because for example, Cell papers generally break up their Results section by headings that summarize the chunk of information pertinent. So, Figure 1 may be looking at the localization of the protein, and the GFP tagged construct, and the first section under Results will be 'Protein Localizes to ER and Vacuoles' or something. As you work through the paper, make a bullet list of information that is conveyed in each figure, and it will provide a rough outline of what the paper is trying to convey. You can then back reference it from the Discussion section, and see if you agree or disagree with what the authors are claiming. Personally, I'm surprised at some of the stuff that makes it into better journals.
... with gigantic melancholies and gigantic mirth, to tread the jeweled thrones of the Earth under his sandalled feet.

User avatar
Angua
Don't call her Delphine.
Posts: 5934
Joined: Tue Sep 16, 2008 12:42 pm UTC
Location: UK/[St. Kitts and] Nevis Occasionally, I migrate to the US for a bit

Re: Reading scientific papers

Postby Angua » Mon Sep 19, 2011 3:22 pm UTC

Thanks - I'll look at that with the figures. Yeah - so far most things have been from lectures, text books, tutorials, and review papers, which generally have a lot of the fine detail about how they actually did the experiment taken out.

For the abstract thing - they blank out the abstract and title when they give you the paper, so parroting it back is not a problem! I doubt he'd ask for an abstract to the ones we're supposed to be reading over the summer though.
Crabtree's bludgeon: “no set of mutually inconsistent observations can exist for which some human intellect cannot conceive a coherent explanation, however complicated”
GNU Terry Pratchett

User avatar
Izawwlgood
WINNING
Posts: 18686
Joined: Mon Nov 19, 2007 3:55 pm UTC
Location: There may be lovelier lovelies...

Re: Reading scientific papers

Postby Izawwlgood » Mon Sep 19, 2011 3:29 pm UTC

Angua wrote:Thanks - I'll look at that with the figures. Yeah - so far most things have been from lectures, text books, tutorials, and review papers, which generally have a lot of the fine detail about how they actually did the experiment taken out.

If you're being quizzed on a paper, my guess is you don't need to know what percentage buffer was used, but you do probably need to know if they did anything new or novel. For example, you can probably say "Up regulation of protein levels was assessed via Western Blotting" but probably don't have to know where they purchased the antibodies from. I would guess. That said, you may need to write about why a particular technique was used; if applicable, you should probably know about limitations of various experiments.
As for the abstract, if you want to practice, post the link to a shortish paper here along with your abstract!
... with gigantic melancholies and gigantic mirth, to tread the jeweled thrones of the Earth under his sandalled feet.

User avatar
Angua
Don't call her Delphine.
Posts: 5934
Joined: Tue Sep 16, 2008 12:42 pm UTC
Location: UK/[St. Kitts and] Nevis Occasionally, I migrate to the US for a bit

Re: Reading scientific papers

Postby Angua » Mon Sep 19, 2011 3:32 pm UTC

Thanks - I might take you up on that when it gets closer to the time!
Crabtree's bludgeon: “no set of mutually inconsistent observations can exist for which some human intellect cannot conceive a coherent explanation, however complicated”
GNU Terry Pratchett

User avatar
LaserGuy
Posts: 4582
Joined: Thu Jan 15, 2009 5:33 pm UTC

Re: Reading scientific papers

Postby LaserGuy » Mon Sep 19, 2011 7:40 pm UTC

Here's what I like to do...

Read the abstract. This will give all of the most important results. Try to get a sense if it is related to what you want to know about. Read the introduction, and see if you can figure out how the paper is positioned and what the important question is. If you already can't figure out what is going on, it's probably a good idea to look at the references in the intro at this point. Maybe run the paper itself, or a few of the references, through Web of Science (or whatever you have) and see how the cross-referencing looks. Probably your paper is no more than 1-2 references away from the "really important" paper on this topic, and that's the one you need to read first to understand what it is all about. Look for papers that lots of the references also cite (again, Web of Science can do this quickly for you). If you can understand the intro, then I'd look at the figures and see what you can make of them (and read their captions, of course). Read the discussion and conclusion. If you sort of think you get it, go back and reread again, maybe skim the parts that you skipped if it seems somewhat important. Then go back to Web of Science, and look at a few related references, or at relevant papers that cited the one that you just read.

User avatar
bigglesworth
I feel like Biggles should have a title
Posts: 7461
Joined: Sat Apr 07, 2007 9:29 pm UTC
Location: Airstrip One

Re: Reading scientific papers

Postby bigglesworth » Mon Sep 19, 2011 7:47 pm UTC

My Developmental Biology professor claimed that all the important information in a biology paper will be in the figures, and that only a short time needs to be spent skimming the rest to find context for the information presented to you in the figures.
Generation Y. I don't remember the First Gulf War, but do remember floppy disks.

User avatar
Izawwlgood
WINNING
Posts: 18686
Joined: Mon Nov 19, 2007 3:55 pm UTC
Location: There may be lovelier lovelies...

Re: Reading scientific papers

Postby Izawwlgood » Mon Sep 19, 2011 8:27 pm UTC

bigglesworth wrote:My Developmental Biology professor claimed that all the important information in a biology paper will be in the figures, and that only a short time needs to be spent skimming the rest to find context for the information presented to you in the figures.

I think the first part of this is true, but unless you are familiar with the field, there's no way in hell you'll understand the material present in the figures in it's entirety. For example, I recently read a paper that utilized both constitutively active and activation null forms of Rac in transfected neurons to look at retrograde transport of a tagged marker. The figure basically read "Wt, RacA19S, RacV12" for those treatments, and you had to find in the Results section what those constructs were.

Basically, the figures will convey what the paper did and is trying to convince you of, but will only include abbreviations and rudimentary explanations of what everything is. Typically, however, each subsection under Results will be tied to a figure or two, and summarize what was found. Like, "Providing trio in MOtor Neurons can partially rescue NMJ defects in BMP mutants", and figure 7 will be labeled "Presynaptic Trio expression can partially... mutants".

Also, I just want to gripe briefly about Drosophila gene names: Please, researchers, stop looking around your attic for ideas on what to name your genes. The pathway I'm working on includes the following gems: Glass Bottom Boat, Mad (for Mothers Against Decapentaplegic), Wishful Thinking, and Saxophone.
... with gigantic melancholies and gigantic mirth, to tread the jeweled thrones of the Earth under his sandalled feet.

qetzal
Posts: 862
Joined: Thu May 01, 2008 12:54 pm UTC

Re: Reading scientific papers

Postby qetzal » Tue Sep 20, 2011 12:26 am UTC

Lots of good advice here, so I'll just add this:

Don't believe what the paper says! It's very possibly wrong - in whole or in part.

A research paper is necessarily 'cutting edge.' It (hopefully) wouldn't get published if it didn't present a good case for its claims, or if it blatantly contradicted well-established science. But even that isn't guaranteed. Remember that the authors are biased in favor of their results being both true and important. Also remember that there's some incentive to play up the importance of the research. (They'll need another grant soon, right?)

The authors' job is to convince you that their findings are (probably) true, and that their overall conclusions follow logically from their specific results. Your job is not to believe them unless they do a good job of that. Of course, it can be really difficult to spot possible holes in their experimentsw or flaws in their logic if you're just learning the field. But you should strive to do so anyway.

If I was your instructor, I'd want to see evidence that you attempted to critically assess the authors' claims for yourself, not just that you can faithfully restate what they claim to have shown. Try to identify the critical experiments in the paper. Then look at what controls the authors included. Try to consider whether their results really mean what they think, or if something else might be going on - especially if it's something they didn't control for. Again, that might be very challenging, especially if you don't know the ins and outs of the techniques, but that's something you should aim for.

User avatar
iop
Posts: 930
Joined: Tue Jul 24, 2007 11:26 am UTC
Location: The ivory tower

Re: Reading scientific papers

Postby iop » Tue Sep 20, 2011 1:10 am UTC

My strategy for reading papers depends a lot on the reason for why I'm reading. If it's just to know what's going on, reading the abstract can be enough (note: this is true in biology. In e.g. physics, the abstract may say "the significance of the results is discussed". What a waste of words, and of my time.). Otherwise, I may read the intro (to learn a bit about the field) and the discussion. I'm not very much a figure person, so I often read without looking at them a lot.

However, there are times when I need to really understand papers from fields with which I'm only tangentially familiar - that's when I review papers. There, I read the paper first from start to end, because that's how the authors intend me to read the text, trying very hard to keep a positive attitude (not easy when the paper is full of spelling mistakes, or rife with wrong statistics!). During this read-through, I make a note on the margin for every claim, and for every bit I don't understand. Then I go back over the parts I didn't understand - maybe the text was just confusing, or maybe I didn't know the method. If it's really important to know the method, I ask the internet, or one of the references (which is sometimes annoying: they reference one of their own paper for the method, and in that paper, they reference another paper by someone else for the method). Then I go over all the claims they made. Are these claims supported by anything presented in the paper? Would I have arrived at a similar claim based on the evidence in the paper, or is it possible to explain this with a much simpler mechanism (this is where the controls are important)? If the conclusions are not supported, is there some re-analysis of the results that could help? If not, what would be a good, easy experiment that the authors could do to make everything work (I hate requesting additional experiments, but sometimes there's just no way around it)?

From there to write a review - or an abstract - it's a very small step: A review goes like this: In this paper, $authors investigate X by methods Y and Z, and they find A, B and C. They then claim that D, which is utter crap (or not). For an abstract, it would be: W is really important. A key mechanism of W is X. Combining the power of Y and Z, we find A and B (say something good about it here). Interestingly, testing some more we find C, which is truly awesome. Together, these findings strongly suggest that D.

MostlyHarmless
Posts: 154
Joined: Sun Oct 22, 2006 4:29 am UTC

Re: Reading scientific papers

Postby MostlyHarmless » Tue Sep 20, 2011 6:02 am UTC

I find it very useful to ask myself, "why did they do/claim that?" for each part of the method/results. If you're somewhat familiar with the field, it's also always good to ask yourself, "why didn't they do/claim this?" This gets back to the previous point about looking for controls, and in particular looking for missed controls.

User avatar
Ulc
Posts: 1301
Joined: Sun Jun 21, 2009 8:05 pm UTC
Location: Copenhagen university

Re: Reading scientific papers

Postby Ulc » Tue Sep 20, 2011 9:10 am UTC

I think there is four levels of reading articles, and they require a different approach.

1: I need a basic understanding of what they did, and why they did it, and the results - but I wont need to cite it, it's just "well, I should keep aware of what is going on in [tangentially related field]".

In this case I'll read the abstract, read the conclusion and skim the figures. Doing that I should have a basic understanding of what was going on, and I'll probably understand it if someone else refers to it, in a "yeah, they did that and the results seemed ok" way.

2: I need a throughout understanding of what they were doing, why they were doing it and the results, but I wont be replicating the experiments or working with the same exact subject.

This starts out pretty much the same. Abstract, then conclusion and skimming the figures. After that they I pen down their most important points as I understood it in 2-3 lines.
After that I'll read it from one end to another, skipping the materials and methods, if I find methods I don't understand, or claims that I don't trust I'll read their references as according to case 1. After that I'll write some more extensive notes - and that's pretty much it, next time I need it, I can just look up the notes and figures.

Quite often I'll be distrustful of results, or methods - for many reasons, but lack of controls, statistical analysis, over-interpretation of data, lack of consideration for other explanations, or overstatement of importance. In this case, I'll look up papers that refer this paper and look for their comments (and if no one has refereed to it, and it's more than 2 years old, I'll pretty much conclude that everyone considered crap).

My tendency to distrust the aper also varies with author, if it's someone I don't know about I'll be more distrustful than if it's a large figure within the field. But that requires quite a bit of knowledge for the field to judge properly.

3: I'll be working with the same subject, and need some of the methods.

As case 2, except I'll be reading the exact parts of the materials and methods I need. If I need their methods on making, say ThT assays, but they're working with a different protein, I'm not going to read the "expression and purification" part.

4: I need some of the methods, but their results aren't particularly relevant to my research or field.

As case 1, except that I read the materials and methods, again picking out what is relevant.


For you, doing it like 2 should probably be sufficient - but make a priority list of the papers, what are most important? Because working your way through a paper in the proper fashion, going back and looking up things, reading the references and making sure you really understand it, tends to take me about an hour per page in the paper. So doing it really carefully becomes impossible quite easily, so consider delegating those that seem least important to the "I need a basic idea about what they did, but no details". And in no cases does it seem like you need materials and methods, no one cares if you know what concentration buffer #27 was.
It is the mark of an educated mind to be able to entertain a thought without accepting it - Aristotle

A White Russian, shades and a bathrobe, what more can you want from life?

User avatar
Izawwlgood
WINNING
Posts: 18686
Joined: Mon Nov 19, 2007 3:55 pm UTC
Location: There may be lovelier lovelies...

Re: Reading scientific papers

Postby Izawwlgood » Tue Sep 20, 2011 1:31 pm UTC

The 'has anyone else referenced this paper' is a good metric for how legit it is. That said, in Biology/Medicine, proper controls are often included side by side with the data in the figures. So if you're skeptical that RNAi of that receptor reduces the signal from it's ligand, lo, they've probably included an image of control transfections with and without ligand.

It's important to be skeptical, to be sure, but for bigger named journals, I think you can dispense with skepticism of their abilities as scientists. I.e., if reading a Cell paper, I shouldn't doubt whether or not they know how to do tissue culture. What I should be skeptical of is their interpretations, because as mentioned in this thread, they want the paper to be accepted, referenced, and adored. So look at the figures and decide for yourself whether or not the result they are claiming appears to be going on, and whether or not it looks like something that is reliably occurring.

Also, take note of the techniques used and perhaps why they were tossed in. I read a paper recently that claimed an effect in retrograde transport, but had a figure entirely dedicated to electrophysiology. I found this a little strange, until my boss, who knew the authors, told me it was because one of the reviewers primarily works in electrophysiology, was unconvinced by the results, and demanded they do a series of marginally related tests. If something seems out of place, it's very possible that it was only included because a reviewer wanted it.
... with gigantic melancholies and gigantic mirth, to tread the jeweled thrones of the Earth under his sandalled feet.

User avatar
Gary_Hurd
Posts: 12
Joined: Tue Jun 21, 2011 3:06 am UTC
Location: California, USA

Re: Reading scientific papers

Postby Gary_Hurd » Tue Sep 20, 2011 7:14 pm UTC

I tried teaching medical students how to use the scientific literature, as opposed to the pharma funded "clinical literature."

I gave up.

It was different for students intent on a research career, whether they were MD, PhD, or combined degree students. So, I decided that clinicians should read clinical papers, and scientists needed to learn how to write clinical papers.

I appologise that this is not helpful to Angua.

For the examination setting you describe, I would suggest reading the paper "backwards." That is, read the conclusions. List them so that you keep them clearly in mind. Now, in the first paragraph, or two, the authors will say why they think their question/topic is important. Add that to the list.

Now, read the paper and see if the reason the paper topic is "important" is related to their conclusions. As you do, check to see what data, or analysis are relevant to which conclusion(s). Does the data, or analysis really relate to the topic, or conclusions? That should do it.


A professional reviewer should pay the strictest attention to the methods sections as there is the greatest likelihood for error. If the methods, or analysis suck, there is little reason to even read the rest of the paper. I also pay a lot of attention to the citations; do the authors mostly cite themselves? If they do, then I am suspicious. If they claim to be "overturning para dimes" then I know they are likely loons.

User avatar
Angua
Don't call her Delphine.
Posts: 5934
Joined: Tue Sep 16, 2008 12:42 pm UTC
Location: UK/[St. Kitts and] Nevis Occasionally, I migrate to the US for a bit

Re: Reading scientific papers

Postby Angua » Wed Sep 21, 2011 9:33 am UTC

Well, the preclinical degree I'm doing at the moment is technically 'Medical Science' and my entire 3rd year is dedicated to research, and that sort of thing, as quite a lot of medics in Oxford end up in research (even if it's alongside their MD careers), so I guess I'm halfway between the two?

Anyway, thanks for all the suggestions - I particularly liked the abstract blueprint, which does seem to fit in well with the abstracts of the papers I'm reading at the moment.

Also, the idea of reading the paper backwards is a good one, I hadn't thought of that. It's certainly interesting to see all the different tactics people take, so I'll definitely be playing around with all of them and seeing what works best for me.
Crabtree's bludgeon: “no set of mutually inconsistent observations can exist for which some human intellect cannot conceive a coherent explanation, however complicated”
GNU Terry Pratchett


Return to “Science”

Who is online

Users browsing this forum: No registered users and 4 guests