Talk on Hodgkin lymphoma cell graphs at ISMB 2016

It’s been a while since I went to the Intelligent Systems for Molecular Biology (ISMB) 2016 conference, but I thought I’d still share a few impressions.



The conference was held at a Disney Resort Hotel near Orlando, FL. Great weather, but not a good place to be as a PhD student who is short on money.


I am working on two projects in the fields of digital pathology/imaging and protein topology in my PhD. I submitted a paper we recently published in Bioinformatics to the highlight track and it was accepted, and I was also given a $ 1000 travel grant (thanks to the RNA group at ISCB!). So I went to the conference directly after a returning from a trip to Indonesia.

The conference was well organized and interesting. There were very few contributions from the field of digital pathology, but quite a few interesting talks on structural biology. I especially enjoyed the keynotes by Ruth Nussinov on Ras signaling and by Deborah Marks on Molecular structure and organism fitness from genomic sequences.


My highlight talk was at the last day of the conference, but I got some good questions afterwards and also met interesting people at the poster sessions that were held earlier during the conference.


If you are interested in the talk, you can find the slides at F1000:

Tim Schäfer, Hendrik Schäfer, Jörg Ackermann, Norbert Dichter, Claudia Döring, Sylvia Hartmann, Martin-Leo Hansmann, and Ina Koch. CD30 cell graphs of Hodgkin lymphoma are not scale-free — an image analysis approach. Highlight talk at Intelligent Systems for Molecular Biology 2016. Orlando, FL, USA

Slide link at F1000:

This is a highlight talk based on the following publication:


Hendrik Schäfer, Tim Schäfer, Jörg Ackermann, Norbert Dichter, Claudia Döring, Sylvia Hartmann, Martin-Leo Hansmann, and Ina Koch. CD30 cell graphs of Hodgkin lymphoma are not scale-free — An image analysis approach. Bioinformatics, 32(1):122–129, 2016. DOI 10.1093/bioinformatics/btv542

Posted in Bioinformatics, general, Science | Tagged , , , , , | Leave a comment

Back from Indonesia

I recently submitted my PhD thesis and then went on a 3 week trip to Sulawesi, Indonesia. We visited Makassar, Rantepao, Tentena, Ampana, the Togean Islands, and Palu. You can find my travel blog with a description of the trip and lots of photos here:

You can find my travel blog with a description of the trip and lots of photos here:

Posted in general | Leave a comment

Elinchrom EL Skyport ECO does not fire flash anymore

I have an Elinchrom D-Lite 2 lighting set in my home studio, and I fire the soft boxes via an Elinchrom EL Skyport ECO transmitter mounted in the camera’s hot shoe.


When I wanted to start a photo session yesterday, the flashes refused to fire. When I clicked the Test button, the LED flashed green, so it seems the batteries were OK. But I didn’t know what else to do (channel settings were good), so I changed the battery. Still no flash.

I searched the internet, and after a while, I found the following line in some product description of the Elinchrom Skyport: Status LED flashes once or twice depending on if Normal-or Speed Sync mode is selected. With my device, it flashed twice.

With this info and new search terms, I finally found this post at

Basically, the Skyport was in its second operation mode (‘speed mode’), and in this mode it does not work with my studio lights. So the solution is to change the mode back to normal. You can do this by holding the Test button until the LED flashes once it seems.

(In the baypattern article it says hold it for 10 seconds, but that did not work in my case.)

Posted in Photography | Tagged , , , , , , | 4 Comments

DSLR filming: oh right, what about the audio?

When you’re a hobbyist DSLR guy like me and try to make your first videos, you will most likely be busy getting the visuals right first. After a while, you may have the idea of turning the sound on when whatching the videos you recorded. At that time, you are most likely be shocked by the cruel audio quality of your oh so great and expensive DSLR!

I thought they gotta be kidding me when I first heard it. My thought was like: I spend € 2000.- on a D800 body, and THIS SHIT is what it gives you for audio?!

The problem

Yes, it’s true. Audio quality of DSLRs sucks hard. Why? Some of the most obvious problems with the built-in DSLR audio include:

  1. Handling noise: The mic is directly attached to the camera without any shockmount (and additionally, it is not directional). So it records the handling noise of the camera, e.g., you pressing buttons / moving the camera, zooming the lens, etc.
  2. Mic quality: The mic most likely has high self noise and/or is not sensitive enough.
  3. Wind (outdoors only): Even very little wind may cause ugly hissing, because the mic has no dead cat or similar protection against wind.
  4. Dependent on camera: You cannot move the mic (independently of the camera). Note that this means you cannot ask somebody to talk directly into the mic — he would not be in the frame of the camera anymore (or be VERY close and look very stupid).

Options to solve the problem

So what can you do to get better audio with a DSLR? This depends on what you want to record, which quality you need, etc. But the general options include:

  • Ignore / circumvent the problem: Decide that audio is not that important or convince yourself that the quality you have is good enough. Or decide to simply not use the recorded audio at all — use background music only for the videos, or whatever. Pros: Dirt cheap, little work! Cons: Crappy or no audio.
  • Buy a video mic: buy a standard, rather cheap (€ 80 — € 200) video mic (like Rode VideoMic, Shure VP83 or similar stuff from other companies). These get mounted in the camera hot shoe, the data is send to your camera for recording via the 3.5 mm jack. Pros: Improves problems 1, 2 and 3 listed above. Rather cheap, medium amount of work (audio is part of the video). Cons: Does not address problem 4. Whether the other problems get solved or only improved depends on your application.
  • Record with an external device: Get a handy recorder (Zoom H4N, Tascam DR-40 or similar stuff from other companies). These are devices with a size between an old cell phone and a Game Boy — you may be able to force them into your pocket, and it is possible to attach them to a DSLR. They have a built-in microphone and they record to internal SD cards these days. Priced between € 180 and € 400. If you don’t go for the cheapest ones, they usually have 2 XLR inputs — this is interesting for high quality audio, because XLR is what professional microphones use! Pros: Possibility to get very good quality. Cons: Expensive, more work (handle 2 devices, sync audio to video later, post-process audio to get even better quality?).

Zoom H5

Actually, the last option could be subdivided further.

  • Record with an external device, but not to it. That is, use the mic of the device, but not the recorder (and storage). Instead, route the audio signal back into the camera for recording (from the monitoring output of the device to the 3.5 mm mic input jack on the camera). Note that this requires a special cable: the mic input on a DSLR requires a mic-level audio signal, and the headphone output of a handy recorder is line-level (see this article by Sam Mallery for details). Pros: Uses good mic of external device, no need to sync audio. Cons: ?
  • Also record to the device. Pros: No need for special cable, can use advanced features of the recording device, e.g., audio filters. Cons: Need to sync audio later

That’s it for now. You now know some options to get better audio when filming with your DSLR.

Posted in audio, Photography | Tagged , , , , , | Leave a comment

Buch-Review: Continuous Delivery — Der pragmatische Einstieg (Wolff)

Note: Most articles on this blog are in English, but I review books in the language they are written in. I read the german edition of this book, so this article is in German.

Das Buch
Title: Continuous Delivery — Der pragmatische Einstieg
Authors: Wolff, E.
Publisher: dpunkt
Year: 2015
Pages: 253

Meine Beschreibung

Das Buch hält, was der Untertitel verspricht, und gibt eine relativ knappe, und sehr praktisch orientierte Einführung in Continous Delivery (CD). CD ist ein Weiterdenken von Continuous Integration (CI), und es geht darum, nach dem Build von Software (wie bei CI) auch die danach folgende Auslieferung so weit wie möglich zu automatisieren. Das Buch richtet sich dabei an unterschiedliche Leserschaften, denn die Einführung von CD betrifft neben den Entwicklern eben auch die Admins und Operations und natürlich das Management. Es werden also auch zu CD passende Organisationsformen aus Nicht-Programmier-Bereichen angesprochen (Lean Startup, DevOps, etc). Tatsächlich geht es eigentlich gar nicht um Software-Entwicklung im Sinne des Programmierens (bis auf die Tests), sondern hauptsächlich um den Aufbau einer CD-Pipeline. Dabei werden sehr pragmatisch dafür nützliche Softwaretools aufgezählt, verglichen, beschrieben und es werden kurze Übungen vorgeschlagen, mit denen man einen guten Eindruck von den Tools bekommt und sie evaluieren kann. Der Fokus liegt im Bereich Open-Source Tools unter Linux/Unix zum Aufbau einer CD-Pipeline für die Entwicklung von Java EE-Anwendungen. Es werden aber auch kommerzielle closed-source Lösungen vorgestellt, und auch die Möglichkeiten zur Nutzung von PaaS (oder Outsourcing der Hardware in die Cloud) werden diskutiert. Zum Aufbau der Pipeline wird dabei viel auf Virtualisierung gesetzt, und es kommen Tools wie Docker und Vagrant zum Einsatz. Build-Tools wie Ant, Maven und Gradle sowie die Automatisierung von Unit-Tests, Akzeptanztests und Lasttests werden ebenfalls praxisnah besprochen. Außerdem unterschiedliche Strategien zum Deployment von Software und Umgang mit Fehlern dabei, inklusive dem dabei heiklen Thema Datenbanken.

… und mein persönlicher Eindruck

Das Buch ist sehr praxisnah und gut lesbar, und die Tiefe, in der die Tools erklärt werden, ist sehr passend gewählt. Ich verstehe mich hauptsächlich als Software-Entwickler, und habe es aus Interesse von vorne bis hinten in recht kurzer Zeit durchgelesen, würde aber sagen, dass es wohl am wertvollsten für die Leute ist, die die Infrastruktur für die CD-Pipeline wirklich einrichten und warten müssten, also in vielen Fällen wohl eher die Administratoren als die Entwickler. Die Praxis-Vorschläge fand ich exzellent. Das Buch richtet sich dabei recht deutlich an Enterprise-Anwender, und manche der vorgestellten Tools und Techniken sind in kleineren Projekten oder Umgebungen (ich wntwickle Software in einem Arbeitskreis an einer dt. Universität) sicher Overkill, aber trotzdem interessant zu kennen. Nach der Lektüre kenne ich die Tools, und kann ihren Nutzen für mich bewerten, und das war das Ziel. Das Buch hat also meine Erwartungen voll erfüllt.

Was ich daraus gelernt habe

  • CD ist ein Weiterdenken von CI, und setzt darauf, alle Schritte von Build und Deployment zu automatisieren. Wichtig sind insbesondere gute Tests.
  • Dies erlaubt einfaches, schnelles und häufiges Deployment. Das führt dazu, dass neu entwickelte Features schnell in Produktion gebracht werden können, und schnell einen Nutzen und Wert bringen.
  • Ich habe viele gute Automatisierungstools kennengelernt (und kürzlich meinen eigenen Jenkins-CI-Server bei OpenShift in Betrieb genommen)
Posted in IT and computers, Software Design | Tagged , , , , , | Leave a comment

Wifi connection slow and unstable on Intel Wireless 7260 and Linux

The problem: choppy and slow wifi with Intel 7260

I recently bought a small 13.3″ notebook, the Lenovo IdeaPad U330p. I formatted the SSD, and installed both Win7 and Lubuntu 15.04 on it (Lubuntu is a light-weight version of the famous Ubuntu, it more or less sacrifices eye-candy for speed). Everything worked perfectly out of the box under Lubuntu it seemed, but when I tried to install new software packages, I noticed that the wifi connection was very slow, it also lost the connection to my router every few minutes and was unusable. The U330p has an Intel 7260 wifi chipset.

My first thought: Damn, it’s like 10 years ago, when I had to fiddle for days with self-compiled kernels and modules which refused to load until I could use the wifi on my laptop under Linux!

Check it

Let’s check the device, to make sure we really have Intel 7260:

# lspci | grep -i wire
02:00.0 Network controller: Intel Corporation Wireless 7260 (rev 73)

OK, we do. Short look at my linux kernel:

$uname -a
Linux mdev 3.19.0-15-generic #15-Ubuntu SMP Thu Apr 16 23:32:37 UTC 2015 x86_64 x86_64 x86_64 GNU/Linux

You can see the connection errors in /var/log/syslog, mine looked like this:

Jun 8 01:03:05 mdev wpa_supplicant[724]: wlan0: WPA: Group rekeying completed with 24:66:11:73:d9:f1 [GTK=TKIP]
Jun 8 01:03:08 mdev whoopsie[664]: [01:03:08] Cannot reach:
Jun 8 01:03:11 mdev kernel: [ 558.707680] wlan0: disassociated from 24:66:11:73:d9:f1 (Reason: 2)
Jun 8 01:03:11 mdev wpa_supplicant[724]: message repeated 6 times: [ wlan0: WPA: Group rekeying completed with 24:65:11:72:d9:f1 [GTK=TKIP]]
Jun 8 01:03:11 mdev wpa_supplicant[724]: wlan0: CTRL-EVENT-DISCONNECTED bssid=24:65:11:72:d9:f1 reason=2
Jun 8 01:03:11 mdev NetworkManager[626]: Connection disconnected (reason 2)

Workaround solution: disable 802.11n on the adapter

So I searched the internet, and found a few topics on this. Some suggested DNS trouble, but the one that did the trick was this great post:

It’s for Ubuntu 14.04, but that did not make a difference. The solution suggested in that post is to disable the 802.11n protocol on the adapter. I had already checked the settings on my router, and it does indeed use 802.11n as the default protocol.

So what is 802.11n?

802.11n is a specification/draft for a faster wifi than the older 802.11g standard. If you disable it, your card will use 802.11g to talk to the access point instead, which is limited to 54 mbit/s. This is more than enough, and most likely more than your internet connection behind the wifi provides. And when using 802.11n with the driver/firmware of Lubuntu, my card has more like 1 mbit/s, so I had nothing to lose. I tried it, and it works for me: after disabling n, the connection is stable and way faster.

So how to disable 802.11n for the Intel 7260 driver in (L)Ubuntu?

This is described by simon in his post I mentioned earlier, and all the credit goes to him:

The solution is the usual way of passing options to drivers in linux, you edit the config file and add the option you want. The driver for the Intel 7260 is called iwlwifi, and the config file is at /etc/modprobe.d/iwlwifi.conf under debian-based distros.

To disable 11n, append the following line to that file:

options iwlwifi 11n_disable=1

Then reboot your laptop, and you should have way better wifi.

Note: In case it does not improve the situation (or in case they improve the driver in the future and you want to give 11n another try then), you may want to re-enable 11n. To do that, just remove the line and reboot once more.

Hope it helps somebody.

Posted in IT and computers | Tagged , , , | 1 Comment

Adware and malware distributed by Sourceforge installers?! (SF) has been a great location to host free and open-source projects for many years, offering services like VCS, forum, bug tracker, etc to open-source developers for free. Many very popular and busy projects are hosted there, and I myself do have some of my projects hosted there.

Malware bundled with FileZilla installer from SF

Recently I downloaded FileZilla, a popular FTP client (they also have a server), from the SF web page. It has been quite a few years since I had one of my machines infected with adware and spyware, but after the FileZilla installation, my browser settings kept changing, ads showed up, I had weird search bars and was transferred to the annoying website every few minutes, and there was no way to remove the software which had hijacked my box and installed all this.

Wow! I had fallen prey to a drive-by installer: an installer for FileZilla, downloaded from Sourceforge, which also installs various junk. Sometimes the junk is just stuff you did not want but can remove easily, but in this case it was some ugly malware that refused to uninstall. I tried some virus scanners, they found stuff, but always something remained. So I had to make a backup of my data and reinstall the OS from scratch.

(I’m not gonna explain why bundling malware sucks, and that a hidden opt-out button is not gonna change this.)

Who’s to blame?

Of course, partly this was my fault: most likely there was a button hidden somewhere in the installer that would have allowed me to not install the malware. So why did this happen to me? Of course, I know stuff like drive-by installers exists on the internet. But: I trusted Sourceforge. I never thought they would use them, and I could not believe it was SF who was to blame. Sadly, some searching on the internet strongly suggests they are indeed to blame:

* This article at arstechnica claims that SF employees taken over the GIMP SF account and have modified the installer of GIMP and other free software on SF and added bundled adware to it:
* This thread at the FileZilla forums suggests that in the case of FileZilla, the FileZilla people know that the installer at SF includes bundled crap (they think it was OK because you can refuse to install it):
* Here is a discussion at reddit about SF shipping adware in installers:

Pressure from GIMP community

* According to this article from The Register published in June 2015, SF has responded to pressure from the GIMP community and removed the installer. The article also mentions a new SF policy and quotes “… we present third party offers only with a few projects where it is explicitly approved by the project developer, or if the project is already bundling third party offers.”

So what?

This still leaves a bad taste for SF. And I just checked out the FileZilla installer at SF, it still has a weird publisher (6 June 2015):

I’m not gonna run that thing again to see whether it installs adware or asks before it does so. And that says it all: my trust in Sourceforge is gone. And I will look into alternatives.

Posted in IT and computers | Tagged , , , , | Leave a comment