Joe Felsenstein talks to Casey Dunn about developing phylogenetic methods

Casey Dunn (Yale EEB) posted a video on his YouTube channel earlier today in which he talks to @Joe_Felsenstein about the development of phylogenetic algorithms and software.

The conversation is a little bit on the technical side if you’re totally unfamiliar with the field, but if you have any interest in phylogenetics or programming, it’s a really fascinating discussion!

2 Likes

Thanks to Casey for that, it worked better than I would have guessed. Warning: almost 2 hours long! It also assumes some knowledge of evolutionary biology and phylogenetic methods. I also am shouting hoarsely at the computer – like most of us old farts I seem to think that to make sure my voice carries all the way through those wires and the network, it is necessary to shout.

3 Likes

It worked really well, I think. I enjoyed both the details regarding the historical context of particular development events and choices and also stepping through the code. As somebody who uses a lot of different scientific software (and writes much less sophisticated code), I always enjoy gaining a clearer picture of what’s going on under the hood.

1 Like

Hey look, some times I just shout at my computer anyway to make it work. I’m still not quite sure what works best, shouting or leaving it be for a few days. Seems to have similar rates of success.

@jeffb ICYMI - seems like this would be of interest

It will be of interest to those who want to know more about the history and workings of phylogeny software. It assumes that the listener has some background in evolutionary biology. Casey intends this to kick off a series of interviews with the people behind the major phylogeny programs.

3 Likes

Curious: did you ever take a look at PHYSYS? I seem to recall that Farris wanted $5000 for a copy.

1 Like

PHYSYS came out around 1982. The cost was that you fly J. S. Farris to your university and have him give a lecture, and you also pay him $5000. He would then bring the source code with him (on a tape or card deck) and you would both go to your computer center where he would have it read in and compiled, then erase the source code and leave you with the executable. So no, no one saw the source code. Later in the 1980s he brought out Hennig86. It was available as an executable for MSDOS and cost $50.

In general I have not spent time trying to look through other people’s source code, as I was too busy trying to work on my own.

3 Likes

Dang. Thank goodness that free, open-source software standards are much more common these days!

4 Likes

It’s been a long time since I did any programming and my understanding of phylogeny is super high-level, so a lot of this stuff is largely outside my understanding—but I will say, those “walked 20 miles in the snow uphill both ways” stories are just amazing.

Well, the charges for most programs in this area were modest. $50 for a phylogeny program when Microsoft Word was selling for $395 or so was not excessive. The people developing the software were not being paid from research grants to do that or to do the work of mailing diskettes. I did have a grant whose granting agency did not seem to object, so I could make my programs free.

3 Likes

I was able to play with PHYSIS briefly as an undergrad, and I thought it sucked, FYI.

… compared to early 1980s versions of PHYLIP and PAUP ?

I’ve never seen either. I think I encountered PHYSYS in 1988, maybe early 1989. First phylogenetic program I ever saw. Now, the only thing I recall is trying out the least squares algorithm, which as I remember just handed you a tree, no claim that it was the optimal tree, no branch swapping, no testing against other trees. Just, apparently, the first tree it happened upon. So I stopped looking at it further.

Farris showed in 1969 that if you set the clustering levels by UPGMA clustering, that for that topology the levels of the clusters were the least squares values. Thus one would get a pretty-good tree that was a least-squares ultrametric tree on the given, approximately-least-squares topology, if one just did UPGMA. Perhaps that’s what it gave you.

I don’t recall that the trees were ultrametric. Also, if you repeated the run you would get a different tree.

1 Like

This topic was automatically closed 7 days after the last reply. New replies are no longer allowed.