About Me
Hello, I’m Thomas Rauter, a bioinformatics scientist who also builds research software. I am doing my PhD in Bioinformatics at the Paris Lodron University of Salzburg, Austria, where I work on the statistical analysis of time-series omics data and on interpretable deep learning for molecular networks.
What I Bring to the Table
My profile combines two skill sets that usually come from different people: the scientific depth to design, carry out, and critically evaluate a scientific analysis, and the software engineering skills to turn that analysis into robust, reusable tools that other scientists can rely on.
Science: Statistical Modeling and Machine Learning
- Linear models and splines: Linear models are the backbone of bioinformatics research, and I worked a lot with them. A key part of my PhD was combining linear models with spline modeling to capture nonlinear changes of molecular features such as transcripts and proteins over time. This work is the basis of my R package SplineOmics which I developed in specific for our research consortium.
- Interpretable neural networks: My second focus is knowledge-primed neural networks (KPNNs), whose architecture mirrors a known molecular network, so every hidden node corresponds to a named gene or protein. This makes it possible to ask not only what a model predicts but which molecular entities it relies on. For this, I developed the Python package kpnn2. A publication on this work is coming soon.
- Broad statistics and machine learning literacy: Beyond these two focus areas, I constantly apply a wide range of statistical and machine learning methods in my work, and I sharpen these skills in data science challenges.
- Causal inference: On my own initiative, I studied causal inference in the framework of Judea Pearl, including causal diagrams (DAGs), the do-operator, and the backdoor criterion. My particular focus was how to apply these methods with linear models, for example to decide which variables a regression must adjust for, and which it must not, so that a coefficient can be read as a causal effect. What fascinates me about the field is that it can estimate the effects of interventions from observational data, where a controlled experiment would be too expensive, unethical, or impossible. I have not yet had the opportunity to apply these methods in my research, but I keep deepening my knowledge, and it shapes how I read any analysis: does it show correlation or causation?
Software: Packages Used by Other Scientists
I developed several R and Python packages that other scientists use in their own research: SplineOmics, scholid, scholidonline, and kpnn2. Three of them are published on CRAN or PyPI. I build them to the standards of production software, with automated tests, continuous integration, and full documentation websites.
Background
I hold a bachelor’s degree in Molecular Biology from Graz and a master’s degree in Biotechnology, where I developed a strong interest in computational methods and statistics. This passion led me to specialize in bioinformatics for my PhD research.
Current Work
As part of my PhD, I:
- Develop statistical methods for time-series omics data, based on linear models and splines.
- Build interpretable neural networks for molecular networks.
- Model CHO cells to improve biotechnological processes.
Strengths
- Autodidactic Learning: I have a proven ability to teach myself complex
technical topics through independent study. With a formal background in
biology, I transitioned into bioinformatics without relying on additional
university coursework. By leveraging online resources, technical
documentation, and hands-on experimentation, I developed expertise in
statistical modeling, data analysis, and machine learning using Python and R. This self-directed approach allows me to efficiently master new tools and concepts in fast-evolving domains. - Structured Thinking: I’m highly organized by nature—whether it’s my desk, desktop, phone, browser tabs, emails, or codebase, everything is intuitively named, neatly arranged, and well-documented. I use to-do lists extensively and rarely lose track of data or files. This structure isn’t just a habit—it’s a conscious strategy. Long-term projects live or die by their organization, and I make sure mine stay alive.
- Patience: Structured work over time requires patience, and I’ve always had a strong sense of that. Progress in research is often slow and incremental, so staying focused on the bigger picture is essential.
- Creativity: I tend to think in unconventional ways, which often leads to efficient or elegant solutions. For example, back in school, I’d sometimes forget the “proper” formulas for math problems but still solve them using graphical reasoning or approximation. That mindset stuck—I look for insight, not just instructions.
Where I am improving
- Prioritization: I sometimes gravitate toward tasks I find interesting, even when they aren’t the most urgent. This can delay higher-priority work. I’ve become more aware of this tendency and am now building routines that help me stay aligned with what matters most to the project—without losing the drive for exploration.
- Focus: I can get distracted by noise and interruptions, especially in shared office environments. To counter this, I use earplugs, limit notifications, and set clear focus blocks to stay fully immersed in one task at a time.
Hardskills
Python programming
Whenever I work with machine learning, Python is my go-to language. I wrote the Python package kpnn2, which builds knowledge-primed neural networks (interpretable neural networks with the architecture of a molecular network) in PyTorch.
R programming
R is the number one language in bioinformatics and therefore I use it a lot in my daily work. I did a multitude of statistical analyses with it, and also wrote the R packages SplineOmics, scholid, and scholidonline.
Softskills
Presenting
Overview:
Presenting results and topics is a key skill when someone pursues a scientific
education. Since highschool, I had to make many different presentations, which
includes more formal ones at seminars and conferences, but also more informal
ones in group meetings. I can say with confidence that over the years I
became a very good presenter, when I have the time to sufficiently prepare.
Explore More
For more details, you can check out: