Dr. Claudio Lottaz studied computer science, mathematics and microelectronics from 1988 to 1994 at the Institute for Computer Science and Applied Mathematics at the University of Bern and the Center Suisse d’Electronique et de Microtechnique at the University of Neuchâtel. Between 1994 and 2000 he was PhD student in computer science at the Artificial Intelligence Laboratory of the Swiss Federal Institute of Technology in Lausanne. In addition, he studied bioinformatics in the master at the Swiss Institute of Bioinformatics from 2000 to 2002. In addition to his studies, he worked at the Swiss Institute for Experimental Cancer Research. From 2002 to 2006 he was a post-doc researcher at the Max Planck Institute for Molecular Genetics in the Department of Bioinformatics.
Since 2007, Mr. Lottaz has been a research assistant and head of research at the Chair for Statistical Bioinformatics at the Institute of Functional Genomics at the Medical Faculty of the University of Regensburg. In his research, Mr. Lottaz deals with modern sequencing technologies.
Research director at the Institute for Functional Genomics
I work as a research assistant at the Chair for Statistical Bioinformatics at the Institute of Functional Genomics in the medical faculty.
I develop exercises, seminars and internships for bioinformatics that we offer in the “Computational Science” course for Bachelor and Master students. As a subsidiary subject, these offers are also attended by physicists, biologists, biochemists and mathematicians.
I am involved in biological and medical prohects in which modern technologies for the high-dimensional measurement of gene, protein or metabolome expressions are used. Usually, enormous amounts of data are generated, which no longer have to be examined by hand, but much more with machine learning methods. Let me mention as an example from the scientific work at the chair an analysis of almost 1000 lymphoma patients. For all these patients, the activity of more than 20.000 genes was measured with routinely impractical effort in order to find 48 gene markers that enable treatment-relevant diagnosis with a cheap and time-efficient technique.
While complex projects and their results can be summarized in lectures, I develop simplified examples for exercises for which students cna find valid results in their own analysis. The fact that the actual benefit of such results is not easy to understand underlines the need for interdisciplinary collaboration in medical biological projects.
For interdisciplinary collaboration, the student learns the necessary vocabulary to be able to discuss with doctors or biologists, to learn from these experts ant to present their own results in an understandable manner. A bioinformatician, on the other hand, is responsible for knowledge of programming, statistics and machine learning. A student can practice all of this in our scientific projects.
More and more scientific techniques enable high-dimensional measurements and are becoming cheaper and cheaper to use. Enormous amounts of data can be accumulated in this way, but their analysis is increasingly challanging classical science. The assessment of data quality and the valid statistical analysis is essential for modern medical and biological projects and the appropriate experts are in short supply.
In the past few years, I have mainly been concerned with modern sequencing technologies. These are increasingly replacing the previous expression measurement methods. As is so often the case with new technologies, quality assurance is a difficult point. One can counteract this problem here in that gene positions are measured several times and a majority vote is then acceptable. The resulting computational effort makes the development of efficient algorithms and clever programming essential in quality assurance.
In my dissertation I dealt with the solution of inequality systems. I developed this in the Artificial Intelligence Lab at the ETH Lausanne. It was about nimerical solutions, which then tried to visualize the solution spaces. At that time, working with architects and civil engineers was the starting point for me to become increasingly interested in interdisciplinary collaboration. After completing my postgraduate studies in bioinformatics, I worked on gene expression analysis, for which I used some machine learning algorithms and developed others.
I studied computer science and worked as an assistant at the University of Bern in the early 1990s, which at that time was considered artificial intelligence. For my master’s thesis I developed a Bayesian network to diagnose thyriod diseases from medical reports. Since I started working as a bioinformatician, I have used machine learning to diagnose diseases or to detect biomarkers.
If we admit that machine learning is part of artificial intelligence, I find it easy to use. The high-dimensional data that appear in all of our projects can no longer be managed in any other way. It is a little more difficult for me to trust the “deep learning” community, which is currently exploring how enormous amounts of data can be used to train enormous neural networks. Some successes cannot be denied, but the complete waiving of the traceability of results is suspect to me.
The strictly targeted science, which first requires a complete hypothesis, then requires a specific measurement and, as a result, only allows confirmation or falsification, may seem too limited in biological and medical science. In high-dimensional data rooms, the number of possible hypotheses is simply too large to find the needle in the haystack, i.e. the correct hypothesis. High-dimensional measuring methods, together with machine learning and artificial intelligence, help to find interesting hypotheses and to later verify them. In science, we have to resist the temptation to accumulate as much data as possible without imaging, in order to be completely surprised by the results.
From 2002 to 2018, I was involved in the large consortium on leukemia and lymphoma. In collaboration with medical professionals, I tried to diagnose patients early to improve treatment decisions. I am currently working loosely with biologists and medical professionals to strengthen our own sequence analysis skills.
I attended my fist lecture on artificial intelligence in 1991 as a student. Back then, this topic still had the charm of science fiction, because existing AI systems could only solve and demonstrate rudimentary tasks. So AI had something playful and was at most threatening in novels. Tody, after 30 years, it is different. AI systems have to be taken seriously. They solve important, serious tasks, are more and more present and also have the potential to become threatening. Today’s students have to learn to understand, use and recognize AI systems in everyday life. This aspect of everyday suitability implies that even people without technical training will have to interact with AI systems. The use of AI on a briad front also implies a mandate for university education to train on a broad front: further development of AI systems for scientists are as well as the recognition of AI in everyday life for everyone to get the AI in check hold.
Networking definitely has the potential to strengthen a location. Both the scientific exchange and the broader education for students is worthwhile. So I am pleased that we can welcome many minor students in our lectures.
AI is currently a huge hype in politics and society. So this questionnaire reads a little one-sidedly. The scientific classification also includes the critical assessment of ethical concerns or control proposals. Of course everyone can draw a purely positive conclusion for themselves, but scientists and teachers also have to think morally.
Mr. Lottaz, thank you for taking the time to answer these questions. We wish you a nice day!