Photo by me
Sorry for not posting lately. I’ve just finished writing and defending my dissertation proposal.
During my first year as a graduate student, James (Jim) Crowe , who is an awesome scientist and a great guy, once told me that I should read at least one journal article everyday. When he first told me this, I wasn’t really sure how to take it. Was he trying to intimidate me, scare me away from science? Or was he pulling one of his prized tips out of his bag, essentially giving me the key to scientific success? Now, in my third year, I would say it was the latter.
Reading one paper every day is tough, for me at least. The articles are dense. One sentence in a scientific article could have taken a grad student like me a year to complete. If you’re reading anything outside your field, you’ll have a tough time simply pronouncing the words correctly. And, if you’re reading about your own field, you still have to become accustomed to new methodologies that are always emerging. It’s not easy, but it gets better with practice.
Keeping up to date on recent scientific publications is essential to be successful in science. All of our work is simply an advancement of the work of others. Furthermore, it’s much easier to formulate a great hypothesis when you are aware of all of the preliminary data, even though this data was generated by another group. Also, if you carefully dissect several high-impact articles, you’ll begin to understand what it takes to publish high-impact papers.
Why read one paper per day? Why not one per week? Why not two per day? Well, one per day seems to be just tough enough that only a few people will do it, and just easy enough to be consistent. A quick glance at Jim Crowe’s pubmed record and the CRISP database would suggest that anything he has to say about science is probably right. The key is to read as much as you can.
I haven’t quite come up with any advice concerning reading review articles as opposed to primary research articles. Maybe someone can shed some light on this in the comments.
Go forth grad students, and read until your eyes burn.
It is for everyone, but it gets easier with time. You get use to the language and the conventions, so you learn which parts are important.
The toughest part is to select which one to read, there are so many on my ‘to read’ list. Any tips?
I think it’s best to stick primarily to the high-impact journals (Nature, Science, Cell, etc.). You can only read a finite number of papers, so why not read the best ones first.
Nuruddeen – I would disagree. Yes, you should read the papers in the top journals, but you’re not going to get a good impression of what’s going on in your field that way – you only see the outliers, which are never representative. The top specialist journals in your field are a better source.
I’d also suggest you read something not in you field occasionally – something that looks interesting or fun. If nothing else, it stops you becoming too set in your ways.
If I were a researcher wanting to keep up with what is going on in my field in the literature, I’d use some combination of customised RSS feed by subject (keyword) alert, together with keeping an eye on PubMed (if discipline covered), Scopus, Google Scholar or other, which provide good services eg most cited/read, similar papers to the one you are looking at.
I think where a “top” journal (mentioning no names) comes into its own as a “regular read” is in giving you a focused, topical, quality overview of the highlights of scientific research across the board, in its review, highlight, and comment sections.
I use:
1) content alerts for top tier journals plus a few lower tier journals of my topic of interest
4) A Yahoo pipe that filters selected journals for keywords related to systems and synthetic biology
2) RSS feeds for Pubmed queries that interest me
3) RSS feeds from Connotea on topics that interest me plus RSS feeeds of people’s libraries that consistently bookmark papers I like
The most useful by far are the Yahoo pipe and the feed for selected people in Connotea since these take the least effort but I like looking for hidden gems so I end up going through a lot of abstracts each week.
The potential counter argument for reading too much is that it could affect your creativity.Here is a thought provoking blog post on this subject by Michael Nielsen
I simply can’t imagine reading only one paper a day or spending only one hour a day reading papers. I don’t consider that I have a good memory, so I have to refer back to papers repeatedly. I do remember the gist of things even if I have spent only a few minutes looking at paper, sometimes even less than a minute. I always look at the references and papers citing the paper I am looking at when I am online downloading it. My default response is to get the paper if it might be useful and also get whatever references and citing articles seem useful (just on the basis of the title and what they seem to be about) and maybe look at them later. I put it into my literature file and then when I am trying to understand a topic I have them to refer back to.
I don’t have an active link to the library I use, so I have to physically go there to download stuff that is not open access. Usually I only do that when I have a new idea I am thinking about and go to get background. Going to the library takes significant time, a lot longer than downloading a paper, so my default is to download stuff if it might be useful. I will think nothing of getting 100+ papers related to the idea, and background on the idea. My preference is to get everything and sort out what is actually useful later. I can never get everything I want; the library I use doesn’t have subscriptions to everything. In no way could I afford paying for everything I want to have and read, so I get what I can and hope that it is “enough”, that there is sufficient redundancy in the literature to satisfy my needs. There is no way I can remember everything either, my hope is that I can remember “enough” to satisfy my needs for as long as I need to remember it.
The biggest “cost” in reading a paper is your time. Being able to manage your time in reading the literature is one of the most important skills you need to be successful. How carefully you read something depends on what you are trying to do with what you are getting from that reading. The answer to the question “when do you know something” is “when you are able to use it”. Modulating how carefully you read something to match what you need to use it for is an extremely important skill. All I need to remember is a few details and that I have the paper in my files. If I need to know more than that I sort through my files until I find it. I could never keep track of everything in hard copy.
You have to look at bad papers as well as good papers so you develop the knowledge and skill to tell them apart. Sometimes it is obvious, sometimes it isn’t. Sometimes part of a paper is ok and another part isn’t. Usually the data is ok, very often the conclusions are not. Being in a high tier journal is no guarantee the paper is good, being in a low tier journal is no guarantee that it is bad. To me, a good paper has good data (i.e. accurate data) that relates to important physiology. A paper with good data and bad conclusions can still be quite useful, much more useful that a paper with bad data. Bad data can be accurate, but just not physiologically relevant. There is a lot of that in the NO field (and in all fields). It is not a surprise that too much NO is bad and kills things. Papers showing 100x physiological levels cause adverse effects are “so what” and nearly useless.
One of the most satisfying feelings for me is having a hypothesis that is not in the mainstream, going to the literature to see what there is about it, finding a paper where they had a different hypothesis, measured something close to what I was interested in, and their data supports my hypothesis and not the one they had and wrote up. What I find so great about that is that the authors published their data even though it didn’t exactly fit their hypothesis. An excellent example is this paper, where they measured ATP levels in humans during septic shock by muscle biopsy. What they found was that ATP levels in septic shock survivors was higher than in matched controls who were not in septic shock, but in non-survivors ATP levels were lower. I think they don’t appreciate the importance of that finding. They don’t even mention it in the abstract (!). To me, it exactly follows my hypothesis of how immune system activation causes mitochondrial failure. At high NO levels mitochondria are turned off by the high ATP level (regulated through sGC) and blocking cytochrome c oxidase. Mitochondria have not “failed” at this point. ATP is then only be supplied by glycolysis. If glycolysis can’t keep up, then ATP falls and mitochondria turn on. To turn on they generate superoxide to pull down the NO level to disinhibit cytochrome c oxidase. Mitochondria have an unlimited capacity to produce superoxide. In the high NO environment of septic shock, that generates too much peroxynitrite in the inner matrix, inhibits MnSOD and kills the mitochondria. Kill too many and you get cell death, kill too many cells and you get organ failure. The “problem” isn’t too much NO, the problem is not enough ATP, which comes from not enough glucose for glycolysis, and/or too much lactate. My idea of how to treat sepsis would be to give as much glucose as possible (until there is hyperglycemia), give insulin until there is insulin resistance (to get cells to activate GLUT transporters to get the glucose inside cells where it is needed), and dialyze with urea containing fluids to get rid of the lactate (so the liver and kidneys don’t activate their mitochondria to keep up with the Cori cycle which would kill them). I recently blogged about this in the context of autism.
I see the literature, even the old literature to be a treasure trove of data that has not been completely understood and explained. If the data is good, it doesn’t matter if it was collected a year ago or 25 years ago. Lately the fad has been to tie everything to gene stuff, which is important, but it can miss a lot that isn’t related to the genetic details. There is a lot of stuff attributed to genes that isn’t actually caused by them (such as autism).
As far as reading review articles and primary articles? Yes, read both. Certainly get both so that if there is a detail in the review article you can go to the primary article it cites. The way that I look at review articles is that they are scientific papers too, but the “data” is the primary literature that they are citing. They are scientific articles about scientific articles. The “conclusions” of the review author(s) may or may not follow from the data they are using, just as the conclusions of an author may not follow from their data. You can’t “understand” a paper except in the right context, that is by understanding the papers cited and which cite it.
Getting “up to speed” in a field is difficult and takes a lot of time and a lot of reading of stuff that you don’t fully understand. The more you read, the more the different things connect together. When you achieve “percolation”, that is when essentially all the ideas become connected in a cluster, is when you become an “expert”. That is a non-linear transition. The percolation threshold is a critical point, where the properties of the network change exponentially with connectivity. An “expert” that has passed the percolation threshold is exponentially more competent than someone who hasn’t. Once your reliable information is connected together in a network, it is easy to see what doesn’t fit and so must be wrong. Eventually this is how you tell the good papers from the bad papers.
It isn’t enough to connect your ideas back to primary papers; you have to connect them back to data. Data is the only thing that really matters. Some of the high tier journals don’t actually have that much data. They certainly don’t have enough data to allow you to connect all the dots yourself which is what you need to do if you are going to really understand the subject yourself.