Correlizer

Beendet
Post Reply
Message
Author
User avatar
odicin
Moderator
Posts: 333
Joined: 25.11.2010, 22:03
Location: Potsdam
Contact:

Correlizer

#1 Post by odicin » 02.08.2011, 21:12

http://svahesrv2.bioquant.uni-heidelberg.de/correlizer/

Team
The sequential organization of genomes, i.e. the relations between distant base pairs and regions within sequences, and its connection to the three-dimensional organization of genomes is still a largely unresolved problem. Long-range power-law correlations were found using correlation analysis on almost the entire observable scale of 132 completely sequenced chromosomes of 0.5 ? 106 to 3.0 ? 107 bp from Archaea, Bacteria, Arabidopsis thaliana, Saccharomyces cerevisiae, Schizosaccharomyces pombe, Drosophila melanogaster, and Homo sapiens. The local correlation coefficients show a species-specific multi-scaling behaviour: close to random correlations on the scale of a few base pairs, a first maximum from 40 to 3,400 bp (for Arabidopsis thaliana and Drosophila melanogaster divided in two submaxima), and often a region of one or more second maxima from 105 to 3 ? 105 bp. Within this multi-scaling behaviour, an additional fine-structure is present and attributable to codon usage in all except the human sequences, where it is related to nucleosomal binding. Computer-generated random sequences assuming a block organization of genomes, the codon usage, and nucleosomal binding explain these results. Mutation by sequence reshuffling destroyed all correlations. Thus, the stability of correlations seems to be evolutionarily tightly controlled and connected to the spatial genome organization, especially on large scales. In summary, genomes show a complex sequential organization related closely to their three-dimensional organization.

aendgraend
Guest
Posts: 16
Joined: 07.04.2009, 09:59

Vorsicht mirt den Tasks von Correlizer

#2 Post by aendgraend » 02.08.2011, 22:19

Vorsicht bei den WUs, die beissen sich mit WUs von AlmereGrid oder deren TestGrid, da beide fast die gleiche Applikation verwenden.

Ausserdem erstellt Correlizer bereits Tasks, die enorm RAM-lastig sind, wenn die WUs noch auf "zur Ausführung bereit" stehen. Ich hatte gerade 1 WU die auf bereit stand, aber schon 14 (!!!) Tasks mit jeweils 240 MB belegtem RAM und 0% CPU-Auslastung hier. Ich habe nun alle AlmereGrid WUs angehalten, nachdem die allesamt nach 1-2 Minuten fehlerhaft beendet wurden.
Jetzt rechnen 2 WUs von Correlizer und haben noch je 1 Task erstellt mit 190 MB RAM-Verbrauch.

Nachtrag:
Es wurden weitere Correlizer-WUs heruntergeladen und schon vorab diese Geister-Tasks erstellt, die alle 240 MB Ram belegen, aber 0% CPU.

Der Rechner hat jetzt 16 Tasks laufen mit je 240 MB RAM und 0% CPU Last, der Name ist correlizer.1.1.0_windows_intelx86.exe (ist eine 32bit Anwendung), und dazu:
5 Tasks die je 190 MB RAM belegen aber komplett unterschiedliche CPU-Last erzeugen, zwischen 8 und 16%, alle 5 laufen, der Name ist correlizer.1.0_windows_intelx86.exe.

Der BOINC Manager zeigt keinen Fortschritt.


Komisches Verhalten, Vorab Geister-Tasks mit viel RAM aber keiner CPU-Last. Mal beobachten...

User avatar
rebirther
Admin
Posts: 2869
Joined: 19.12.2005, 00:59
Location: Germany

#3 Post by rebirther » 03.08.2011, 19:03

Sieht besser aus als AlmereGrid. Die WU läuft mit 100% Last und verbraucht 190MB RAM (win7 x64)

User avatar
odicin
Moderator
Posts: 333
Joined: 25.11.2010, 22:03
Location: Potsdam
Contact:

#4 Post by odicin » 04.08.2011, 16:46

Jupp, einzig der Boinc Manager zeigt den Fortschritt nicht an. Mit Boinctasks wird auch der aktuelle Fortschritt korrekt angezeigt.

Gruß Odi

User avatar
odicin
Moderator
Posts: 333
Joined: 25.11.2010, 22:03
Location: Potsdam
Contact:

#5 Post by odicin » 09.08.2011, 07:26

Hier übrigens die Erklärung von Honto ni dazu, das bei Correlizer und Almere Grid die gleichen Anwendungen laufen:
About the collaboration with other grids, the work on the dedicated clusters and the Erasmus Computing Grid all use the same sequence but calculate different window sizes. For the BOINC infrastructures this is currently not the case, here each server calculates another sequence but in all cases no duplicated work is send (except of course for the redundant workunits). This will change in the near future when the bridge is completely hooked up with the backend, then it will also work on the same sequence as the other grids do.
Gruß Odi

mislaura7

#6 Post by mislaura7 » 29.01.2015, 08:18

I can't emphasize enough how critical all of your contributions are to our efforts to improve protein structure prediction methods! we have had many ideas on potential improvements which we have implemented, but the only way to know whether we are actually getting closer to the truth is to do rigorous testing (the casp7 targets for which we now know the structures are a perfect test set), and this would quite simply be impossible without your contributions! --

User avatar
KW5
BC-Member
Posts: 35
Joined: 15.09.2013, 22:21
Location: USA

#7 Post by KW5 » 29.01.2015, 09:12

@ mislaura7

Roger,confirmed, I`m in !

User avatar
odicin
Moderator
Posts: 333
Joined: 25.11.2010, 22:03
Location: Potsdam
Contact:

Re: Correlizer

#8 Post by odicin » 24.05.2020, 17:53

Moved to retired...

Post Reply