In the year 2022, the Nobel Prize in Medicine was awarded to Svante Pääbo for his discoveries concerning the genomes of extinct hominins and human evolution. Some 25 years earlier, among other things, the famous TREE-PUZZLE phylogeny software had been developed at Pääbo's institute, which at that time was in Munich. The software had been developed by Korbinian Strimmer and Arndt von Haeseler and was published here:
In those early days of phylogeny softwares I did my diploma thesis under the supervision of Arndt von Haeseler on some unimportant topic using TREE-PUZZLE, but along the way I found it much more interesting to recode and improve the TREE-PUZZLE program, and I soon had created a phylogeny program of my own. Not yet a really good one, but this is how the TREEFINDER project began.
After I had already spent a lot of work on my diploma thesis and also on my early TREEFINDER program, Pääbo suddenly decided for purely egoistic reasons that his whole institute and everyone working there would have to move from Munich to Leipzig. It was only for money. Hardly anyone liked the idea, and quite a many good people have therefore left the institute. I, too, did not want to move to Leipzig, but I somehow managed to finish my diploma thesis and then was left alone in Munich.
When Pääbo's institute had moved to Leipzig, I could not immediately continue my TREEFINDER project as a PhD thesis with Arndt von Haeseler, as I originally had planned. But while looking for some alternative job in Munich, I still continued improving TREEFINDER, hoping that this work might be helpful some day. I tried several new tree search algorithms of my own and I built a graphical user interface - and, most of all, I did a lot of testing. I tested the same way as TREE-PUZZLE had been tested in its publication. I did the simulations using my newly created programming language TL, which made everything much easier. After a few months, I finally managed to outperform the quartet puzzling method with one of my own algorithms in terms of the percentage of correctly reconstructed trees.
While testing and comparing my program with others I noticed that TREE-PUZZLE did not at all perform as described in its paper. I checked everything several times, but the results I got were always worse than the ones published. I was not yet sure what that meant, but I was glad that my own algorithm performed well enough, so I wrote an email to Arndt in Leipzig an told him the good news.
Then Arndt suddenly became very interested in my program and he invited me to Leipzig, to Pääbo's newly founded Max-Planck-Institute for Evolutionary Anthropology. There we tested TREEFINDER and we compared it with TREE-PUZZLE and also with Felsenstein's PHYLIP program, and the simulations took quite a while. When the simulations were finally finished, we viewed the results and I wondered why Arndt was not as happy about them as I was. "Why is TREEFINDER that good?", he asked displeased.
But Arndt offered me a PhD position in his group and, best of all, this position was without the necessity for me to move to Leipzig. The agreement was that I would have to do most of my work at home in Munich and would come visiting him in Leipzig every second week. But, unfortunately, I would have to pay a considerable part of my low salary for train tickets.
However, the results we obtained with TREE-PUZZLE were much worse than the results published in its paper. TREE-PUZZLE does not hold what is promised. The following tables show the differences between the published and the recomputed percentages of correctly reconstructed trees according to the test from the TREE-PUZZLE paper. Except that I did 10,000 repetitions, just to be sure.
| published: | 71.5 | 54.4 | 11.3 | 93.8 | 86.0 | 36.6 | 57.8 | 42.5 | 14.2 | 87.0 | 75.3 | 35.6 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| recomputed: | 46.3 | 26.9 | 0.8 | 86.2 | 71.2 | 6.8 | 31.6 | 17.7 | 1.4 | 72.9 | 56.1 | 10.6 |
| published: | 83.6 | 75.3 | 33.3 | 96.7 | 93.5 | 59.2 | 74.2 | 65.8 | 36.5 | 94.8 | 88.4 | 61.7 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| recomputed: | 70.4 | 64.6 | 20.8 | 94.9 | 92.1 | 49.3 | 59.5 | 53.8 | 22.2 | 90.0 | 85.3 | 51.7 |
A TL script for repeating the TREE-PUZZLE test as described in the paper can be downloaded here:
After I had worked a few months at the Max-Planck-Institute for Evolutionary Anthropology, it turned out that Arndt did not truely intend to award me a doctorate some day, but was only interested in getting hands on my program code. In a talk at the institute he suddenly announced that parts of my TREEFINDER program code would soon be incorporated into his TREE-PUZZLE program - but we had previously never discussed that. In my talk, which was after Arndts talk, I informed the audience that my TREEFINDER program code would not be incorporated into the TREE-PUZZLE program and that Arndt was wrong - and Arndt and the colleagues did only laugh at me. But in the end I was right: I resigned from that job.
At the end of the year 2022 I am alone and poor and I must earn my money as an unskilled worker in a bakery. I cannot work as a well-paid scientific programmer because the science system that Pääbo represents keeps me out. The Pääbo system must be overcome.
I do not question the importance of Pääbo's discoveries, but his research has come along with too much social collateral damage and it is not that worth. The moving of an institute from one city to another means the disruption of many people's lives, carreers, families. The miserable salaries paid at his institute do not enable young researchers to buy an own house or to to reasonably found a family, makes them waste their valuable life times on futureless carreers. Pääbo's style of research promotes migration and carreer nomadism, the alienation from one's home place, nation, culture. Pääbo's research style promotes globalism, is therefore harmful to mankind and should be not awarded with any prize at all.
After several years with no software update I see that TREEFINDER is still being downloaded and used, and I see that my retracted paper is still being cited. I can imagine that some people may want the TREEFINDER software to be updated and modernized, and I can imagine that some people may want to support me. Either to produce a new TREEFINDER version in the future, or as a compensation for my work in the past, depending on how much support I will get. I suggest the following three methods to support me:
First, I would appreciate donations of money.
Second, I would appreciate a competitive salary from whereever for producing new TREEFINDER versions, but I would do this here at my home.
Third, I may be willing to sell the TREEFINDER project together with all of the source codes to some company interested in doing the further development.
The TREEFINDER license conditions are negotiable. For donators from sanctioned countries I can make license exceptions.
Please send me your offers to the email address shown in the TREEFINDER manual.