Lior Pachter

I read abstracts of articles on the @biorxivpreprint every morning and every evening, and typically a few times during the day as well. I am sometimes delving into the main paper, usually starting with the supplement. 2/

I search for articles on @Google using keywords (and filtering for "past 24 hours"), my lab has @SlackHQ channels where people post articles, I check @rxivist, and I read and search twitter by keywords for topics I'm interested in every day. I follow @biorxivpreprint. 3/

I find that @Twitter can be very biased, promoting preprints on the basis of popularity of the authors. I try to consciously filter for such bias, specifically seeking out preprint tweets from individuals with few(er) followers who I don't personally know. 5/

I have no complaints about the preprints my group recommends on our @SlackHQ. They (usually) have great taste :) 6/

I start the process of reading a preprint with the supplement. I begin with a few basic quick checks to establish a prior about the quality the work that I'm looking at. 7/

I check for a data availability statement, and click on the link(s) to see that the data actually exists where claimed. Many preprints don't even have links to data. If I see that going on my prior shifts towards "junk". 8/

I check for the software. I try to find code that reproduces all the results and figures in the preprint. I do a quick check of the license, the programming language, and I will occasionally download and see if I can get it up and running quickly (I try for a few min. max). 10/

Speaking of being familiar with fields, I search every day in all the sites I mentioned for terms including #rnaseq, #scRNAseq, #genomics, #statgen, #popgen, #phylogenetics, #neuroscience, #combinatorics, and recently... #jellyfish. Sometimes other things. 13/

I probably read the @biorxivpreprint a few hours a day. A lot of it happens on my phone between meetings, downtime in various places, etc. I do have to admit sometimes I procrastinate this way. 14/

Weird non-sensical p-values are a common problem in tables. But one can find a lot of dirty laundry in there on quick inspection. BTW same goes for figures. https://www.biostat.wisc.edu/~kbroman/topten_worstgraphs/ … 16/

I find a common problem nowadays is people performing multiple-testing correction for their p-values with one multiple testing correction per test. Then the authors report dozens of p-values from different experiments and argue that one of them is significant. 18/

Now here is one useful trick I learned from a collaborator: I do a quick search for "the the". It's a common typo that is caught if a preprint is properly proofread. There is a surprising correlation between failing "the the" test and preprints even the authors haven't read. 21/

With a prior at hand l may delve into the preprint. The point of the prior is that it guides me when I'm stuck. If my prior is high I doubt myself and try again. If my prior is low, I wonder whether the problem is with the preprint, not me. This reduces imposter syndrome. 22/

There is so much material on the @biorxivpreprint now that I can't keep track of most of it anymore. A few years ago I was doing better. The #singlecellRNAseq alone is overwhelming, although focusing on methods papers helps. Still, I know I'm missing lots of good stuff. 23/

Finally, if I like a preprint I print it out on paper. I've found that reading off paper is really helpful and improves my memory retention. Protip: when *writing* a paper read it *out loud* off a hardcopy to improve the writing. 24/