March 25, 2016 at 04:29AM

Today I Learned: 1) ...that controlled fires are an important tool for management of the Florida Everglades. In addition to the usual benefits of controlled fires (recycling of nutrients, clearance of possible uncontrolled fire triggers, etc), fire is particularly useful in the Everglades for clearing out invasive species from cyprus groves -- the cyprus trees are *much* more resilient against fire than any invasive plants, so fire really does purify the groves. Thanks Mengsha Gong! 2) J. Craig Venter (or, more accurately, the J. Craig Venter Institute) has made a new minimal synthetic genome! It's a stripped-down version of Mycoplasm mycoides, which is the same species JCVI used in their last minimal genome paper two years ago. M. mycoides is a good candidate for building a minimal genome because it's already quite small (901 genes). This is probably because M. mycoides is parasitic, and doesn't synthesize a lot of its own metabolites. In any case, the newest version of M. mycoides, dubbed syn3.0 by the research team, has 473 genes. Of those, 149 have unknown function but are required for growth. It has a doubling time of about three hours, which is slow, but not *absurdly* slow -- for reference, that puts it on par with many cyanobacteria for replication rate. A number of people independently alerted me to this story today, including Science Magazine. Primary article at http://ift.tt/1UMC0yg, but it's Science so it's almost certainly paywalled unless you're at an academic institution. Science also has a review article here (http://ift.tt/1T7YsRo) that should be free -- please let me know if it isn't and I'll find a better one. 3) ...about the Credence Calibration Game (see http://ift.tt/1dzDceM). The idea of the game is to train yourself to correctly estimate your confidence on statements. It goes like this. The game asks you a question, like "which country has the higher GDP, France or Saudi Arabia?". You give an answer, and a confidence (50%, 60%, 70%, 80%, 90%, or 99%). If you guess right, you get points; if you guess wrong, you lose points. The higher your reported confidence, the more points you get/lose. Theoretically, the point values are calibrated so that you do best if you correctly estimate your confidence (i.e., you should be getting 50% of your 50% confidence answers correct, 60% of your 60% confidence answers correct, etc). I've seen this game in a couple of forms, but this one's nice because it's a discrete, downloadable app that's totally automated (no tallying points yoruself) and a tight gameplay loop. Go check it out if you're interested in improving your own probability estimation performance!