arXiv subjects

Alexander Schönhuth

Publications and source records attributed to Alexander Schönhuth.

18 recordsLinked to original sources

arXiv2026

CancerZigZag: Iterative Seed-Anchored Diffusion for Generative Modeling of Single-Cell State Transitions

Single-cell cancer datasets are predominantly cross-sectional and rarely provide paired or longitudinal observations linking individual healthy-like cells to tumor-associated states. We introduce CancerZigZag, a seed-initialized diffusion-based framework for exploratory generation of tumor-associated single-cell candidate clouds from unpaired epithelial cell populations. For each cancer context, a diffusion model is trained exclusively on tumor-derived epithelial cells and applied through repeated partial latent-space perturbation and reverse diffusion to held-out healthy-like seeds, generating stochastic candidate clouds without paired measurements or classifier guidance. We applied CancerZigZag to colorectal, breast, lung, and renal cell carcinoma contexts and explored parameter landscapes defined by perturbation depth and the number of ZigZag cycles. Across the reported operating configurations, candidate clouds contained outputs classified toward held-out tumor-derived reference populations for each evaluated seed. Residual seed-dependent organization varied across contexts, with the clearest structure in colorectal cancer, more modest organization in lung cancer, and limited cloud-level structure in breast and renal cell carcinoma. Representative candidates also showed directional concordance with transcriptional shifts observed between held-out healthy-like and tumor-derived reference populations. CancerZigZag is not interpreted as a model of deterministic healthy-to-tumor transformation or cellular progression. Instead, it provides a reference-informed framework for exploring tumor-associated candidate distributions from unpaired healthy-like seeds and quantifying the context-dependent relationship between tumor-associated displacement and residual seed dependence.

q-bio.GN↗

arXiv2026

Retracing the Process of Translation: Proteome-wide mapping of stable transcriptomic predictors of protein abundance in cancer cell lines

Understanding the relationship between gene expression and protein abundance is central to molecular and systems biology. While gene expression reflects transcriptional activity, proteins are the functional molecules that determine cellular phenotypes. However, numerous post-transcriptional and translational regulatory layers complicate this relationship, and prior studies have reported only weak to moderate correlations between RNA and protein levels. Predicting protein abundance from transcriptomic data remains challenging, but it is a valuable goal for biological insight, especially when proteomic data is limited or unavailable. In this study, we applied a large-scale, Ridge regression-based feature selection strategy to identify predictive gene expression features for each of 8,423 proteins across 940 cancer cell lines. To our knowledge, this is the first work to perform such comprehensive protein-wise feature selection at this scale. Our analysis revealed both globally predictive and context-specific gene features. These included biologically meaningful modules such as immune-related genes, HOX transcription factor targets, and cytoskeletal components. The models identified stable candidate gene-protein associations that remained interpretable at the level of individual proteins and recurrent transcriptomic predictor patterns. Our approach enables interpretable modeling of protein expression from transcriptomic data and provides insight into transcriptomic features associated with protein abundance. This framework may support hypothesis generation, protein imputation in incomplete datasets, and deeper understanding of post-transcriptional regulation in cancer biology.

q-bio.QM↗

arXiv2025

Genetic Features for Drug Responses in Cancer -- Investigating an Ensemble-Feature-Selection Approach

Predicting drug responses using genetic and transcriptomic features is crucial for enhancing personalized medicine. In this study, we implemented an ensemble of machine learning algorithms to analyze the correlation between genetic and transcriptomic features of cancer cell lines and IC50 values, a reliable metric for drug efficacy. Our analysis involved a reduction of the feature set from an original pool of 38,977 features, demonstrating a strong linear relationship between genetic features and drug responses across various algorithms, including SVR, Linear Regression, and Ridge Regression. Notably, copy number variations (CNVs) emerged as more predictive than mutations, suggesting a significant reevaluation of biomarkers for drug response prediction. Through rigorous statistical methods, we identified a highly reduced set of 421 critical features. This set offers a novel perspective that contrasts with traditional cancer driver genes, underscoring the potential for these biomarkers in designing targeted therapies. Furthermore, our findings advocate for IC50 values as a predictable measurement of drug responses and underscore the need for more data that can represent the dimensionality of genomic data in drug response prediction. Future work will aim to expand the dataset and refine feature selection to enhance the generalizability of the predictive model in clinical settings.

q-bio.GN↗

arXiv2025

Generating Synthetic Genotypes using Diffusion Models

In this paper, we introduce the first diffusion model designed to generate complete synthetic human genotypes, which, by standard protocols, one can straightforwardly expand into full-length, DNA-level genomes. The synthetic genotypes mimic real human genotypes without just reproducing known genotypes, in terms of approved metrics. When training biomedically relevant classifiers with synthetic genotypes, accuracy is near-identical to the accuracy achieved when training classifiers with real data. We further demonstrate that augmenting small amounts of real with synthetically generated genotypes drastically improves performance rates. This addresses a significant challenge in translational human genetics: real human genotypes, although emerging in large volumes from genome wide association studies, are sensitive private data, which limits their public availability. Therefore, the integration of additional, insensitive data when striving for rapid sharing of biomedical knowledge of public interest appears imperative.

cs.CE↗

arXiv2020

Predicting potential drug targets and repurposable drugs for COVID-19 via a deep generative model for graphs

Coronavirus Disease 2019 (COVID-19) has been creating a worldwide pandemic situation. Repurposing drugs, already shown to be free of harmful side effects, for the treatment of COVID-19 patients is an important option in launching novel therapeutic strategies. Therefore, reliable molecule interaction data are a crucial basis, where drug-/protein-protein interaction networks establish invaluable, year-long carefully curated data resources. However, these resources have not yet been systematically exploited using high-performance artificial intelligence approaches. Here, we combine three networks, two of which are year-long curated, and one of which, on SARS-CoV-2-human host-virus protein interactions, was published only most recently (30th of April 2020), raising a novel network that puts drugs, human and virus proteins into mutual context. We apply Variational Graph AutoEncoders (VGAEs), representing most advanced deep learning based methodology for the analysis of data that are subject to network constraints. Reliable simulations confirm that we operate at utmost accuracy in terms of predicting missing links. We then predict hitherto unknown links between drugs and human proteins against which virus proteins preferably bind. The corresponding therapeutic agents present splendid starting points for exploring novel host-directed therapy (HDT) options.

q-bio.MN↗

arXiv2019

A Drug Recommendation System (Dr.S) for cancer cell lines

Personalizing drug prescriptions in cancer care based on genomic information requires associating genomic markers with treatment effects. This is an unsolved challenge requiring genomic patient data in yet unavailable volumes as well as appropriate quantitative methods. We attempt to solve this challenge for an experimental proxy for which sufficient data is available: 42 drugs tested on 1018 cancer cell lines. Our goal is to develop a method to identify the drug that is most promising based on a cell line's genomic information. For this, we need to identify for each drug the machine learning method, choice of hyperparameters and genomic features for optimal predictive performance. We extensively compare combinations of gene sets (both curated and random), genetic features, and machine learning algorithms for all 42 drugs. For each drug, the best performing combination (considering only the curated gene sets) is selected. We use these top model parameters for each drug to build and demonstrate a Drug Recommendation System (Dr.S). Insights resulting from this analysis are formulated as best practices for developing drug recommendation systems. The complete software system, called the Cell Line Analyzer, is written in Python and available on github.

stat.ML↗

arXiv2018

An image representation based convolutional network for DNA classification

The folding structure of the DNA molecule combined with helper molecules, also referred to as the chromatin, is highly relevant for the functional properties of DNA. The chromatin structure is largely determined by the underlying primary DNA sequence, though the interaction is not yet fully understood. In this paper we develop a convolutional neural network that takes an image-representation of primary DNA sequence as its input, and predicts key determinants of chromatin structure. The method is developed such that it is capable of detecting interactions between distal elements in the DNA sequence, which are known to be highly relevant. Our experiments show that the method outperforms several existing methods both in terms of prediction accuracy and training time.

cs.LG↗

arXiv2018

Dualities in Tree Representations

A characterization of the tree $T^*$ such that $\mathrm{BP}(T^*)=\overleftrightarrow{\mathrm{DFUDS}(T)}$, the reversal of $\mathrm{DFUDS}(T)$ is given. An immediate consequence is a rigorous characterization of the tree $\hat{T}$ such that $\mathrm{BP}(\hat{T})=\mathrm{DFUDS}(T)$. In summary, $\mathrm{BP}$ and $\mathrm{DFUDS}$ are unified within an encompassing framework, which might have the potential to imply future simplifications with regard to queries in $\mathrm{BP}$ and/or $\mathrm{DFUDS}$. Immediate benefits displayed here are to identify so far unnoted commonalities in most recent work on the Range Minimum Query problem, and to provide improvements for the Minimum Length Interval Query problem.

cs.DS↗

arXiv2015

On Hidden States in Quantum Random Walks

It was recently pointed out that identifiability of quantum random walks and hidden Markov processes underlie the same principles. This analogy immediately raises questions on the existence of hidden states also in quantum random walks and their relationship with earlier debates on hidden states in quantum mechanics. The overarching insight was that not only hidden Markov processes, but also quantum random walks are finitary processes. Since finitary processes enjoy nice asymptotic properties, this also encourages to further investigate the asymptotic properties of quantum random walks. Here, answers to all these questions are given. Quantum random walks, hidden Markov processes and finitary processes are put into a unifying model context. In this context, quantum random walks are seen to not only enjoy nice ergodic properties in general, but also intuitive quantum-style asymptotic properties. It is also pointed out how hidden states arising from our framework relate to hidden states in earlier, prominent treatments on topics such as the EPR paradoxon or Bell's inequalities.

quant-ph↗

arXiv2013

Generic identification of binary-valued hidden Markov processes

The generic identification problem is to decide whether a stochastic process $(X_t)$ is a hidden Markov process and if yes to infer its parameters for all but a subset of parametrizations that form a lower-dimensional subvariety in parameter space. Partial answers so far available depend on extra assumptions on the processes, which are usually centered around stationarity. Here we present a general solution for binary-valued hidden Markov processes. Our approach is rooted in algebraic statistics hence it is geometric in nature. We find that the algebraic varieties associated with the probability distributions of binary-valued hidden Markov processes are zero sets of determinantal equations which draws a connection to well-studied objects from algebra. As a consequence, our solution allows for algorithmic implementation based on elementary (linear) algebraic routines.

math.ST↗

arXiv2013

Next Generation Cluster Editing

This work aims at improving the quality of structural variant prediction from the mapped reads of a sequenced genome. We suggest a new model based on cluster editing in weighted graphs and introduce a new heuristic algorithm that allows to solve this problem quickly and with a good approximation on the huge graphs that arise from biological datasets.

cs.DS↗

arXiv2013

Sensitive Long-Indel-Aware Alignment of Sequencing Reads

The tremdendous advances in high-throughput sequencing technologies have made population-scale sequencing as performed in the 1000 Genomes project and the Genome of the Netherlands project possible. Next-generation sequencing has allowed genom-wide discovery of variations beyond single-nucleotide polymorphisms (SNPs), in particular of structural variations (SVs) like deletions, insertions, duplications, translocations, inversions, and even more complex rearrangements. Here, we design a read aligner with special emphasis on the following properties: (1) high sensitivity, i.e. find all (reasonable) alignments; (2) ability to find (long) indels; (3) statistically sound alignment scores; and (4) runtime fast enough to be applied to whole genome data. We compare performance to BWA, bowtie2, stampy and find that our methods is especially advantageous on reads containing larger indels.

q-bio.GN↗

arXiv2012

CLEVER: Clique-Enumerating Variant Finder

Next-generation sequencing techniques have facilitated a large scale analysis of human genetic variation. Despite the advances in sequencing speeds, the computational discovery of structural variants is not yet standard. It is likely that many variants have remained undiscovered in most sequenced individuals. Here we present a novel internal segment size based approach, which organizes all, including also concordant reads into a read alignment graph where max-cliques represent maximal contradiction-free groups of alignments. A specifically engineered algorithm then enumerates all max-cliques and statistically evaluates them for their potential to reflect insertions or deletions (indels). For the first time in the literature, we compare a large range of state-of-the-art approaches using simulated Illumina reads from a fully annotated genome and present various relevant performance statistics. We achieve superior performance rates in particular on indels of sizes 20--100, which have been exposed as a current major challenge in the SV discovery literature and where prior insert size based approaches have limitations. In that size range, we outperform even split read aligners. We achieve good results also on real data where we make a substantial amount of correct predictions as the only tool, which complement the predictions of split-read aligners. CLEVER is open source (GPL) and available from http://clever-sv.googlecode.com.

q-bio.GN↗

arXiv2012

A Markovian Model for Joint Observations, Bell's Inequality and Hidden States

While the standard approach to quantum systems studies length preserving linear transformations of wave functions, the Markov picture focuses on trace preserving operators on the space of Hermitian (self-adjoint) matrices. The Markov approach extends the standard one and provides a refined analysis of measurements and quantum Markov chains. In particular, Bell's inequality becomes structurally clear. It turns out that hidden state models are natural in the Markov context. In particular, a violation of Bell's inequality is seen to be compatible with the existence of hidden states. The Markov model moreover clarifies the role of the "negative probabilities" in Feynman's analysis of the EPR paradox.

cs.IT↗

arXiv2011

Mirroring co-evolving trees in the light of their topologies

Determining the interaction partners among protein/domain families poses hard computational problems, in particular in the presence of paralogous proteins. Available approaches aim to identify interaction partners among protein/domain families through maximizing the similarity between trimmed versions of their phylogenetic trees. Since maximization of any natural similarity score is computationally difficult, many approaches employ heuristics to maximize the distance matrices corresponding to the tree topologies in question. In this paper we devise an efficient deterministic algorithm which directly maximizes the similarity between two leaf labeled trees with edge lengths, obtaining a score-optimal alignment of the two trees in question. Our algorithm is significantly faster than those methods based on distance matrix comparison: 1 minute on a single processor vs. 730 hours on a supercomputer. Furthermore we have advantages over the current state-of-the-art heuristic search approach in terms of precision as well as a recently suggested overall performance measure for mirrortree approaches, while incurring only acceptable losses in recall. A C implementation of the method demonstrated in this paper is available at http://compbio.cs.sfu.ca/mirrort.htm

q-bio.PE↗

arXiv2010

Efficient tests for equivalence of hidden Markov processes and quantum random walks

While two hidden Markov process (HMP) resp. quantum random walk (QRW) parametrizations can differ from one another, the stochastic processes arising from them can be equivalent. Here a polynomial-time algorithm is presented which can determine equivalence of two HMP parametrizations $\cM_1,\cM_2$ resp. two QRW parametrizations $\cQ_1,\cQ_2$ in time $O(|§|\max(N_1,N_2)^{4})$, where $N_1,N_2$ are the number of hidden states in $\cM_1,\cM_2$ resp. the dimension of the state spaces associated with $\cQ_1,\cQ_2$, and $§$ is the set of output symbols. Previously available algorithms for testing equivalence of HMPs were exponential in the number of hidden states. In case of QRWs, algorithms for testing equivalence had not yet been presented. The core subroutines of this algorithm can also be used to efficiently test hidden Markov processes and quantum random walks for ergodicity.

cs.IT↗

arXiv2010

Pair HMM based gap statistics for re-evaluation of indels in alignments with affine gap penalties: Extended Version

Although computationally aligning sequence is a crucial step in the vast majority of comparative genomics studies our understanding of alignment biases still needs to be improved. To infer true structural or homologous regions computational alignments need further evaluation. It has been shown that the accuracy of aligned positions can drop substantially in particular around gaps. Here we focus on re-evaluation of score-based alignments with affine gap penalty costs. We exploit their relationships with pair hidden Markov models and develop efficient algorithms by which to identify gaps which are significant in terms of length and multiplicity. We evaluate our statistics with respect to the well-established structural alignments from SABmark and find that indel reliability substantially increases with their significance in particular in worst-case twilight zone alignments. This points out that our statistics can reliably complement other methods which mostly focus on the reliability of match positions.

q-bio.QM↗

arXiv2009

On analytic properties of entropy rate

Entropy rate is a real valued functional on the space of discrete random sources which lacks a closed formula even for subclasses of sources which have intuitive parameterizations. A good way to overcome this problem is to examine its analytic properties relative to some reasonable topology. A canonical choice of a topology is that of the norm of total variation as it immediately arises with the idea of a discrete random source as a probability measure on sequence space. It is shown that entropy rate is Lipschitzian relative to this topology, which, by well known facts, is close to differentiability. An application of this theorem leads to a simple and elementary proof of the existence of entropy rate of random sources with finite evolution dimension. This class of sources encompasses arbitrary hidden Markov sources and quantum random walks.

cs.IT↗