arXiv subjects

Niko Beerenwinkel

Publications and source records attributed to Niko Beerenwinkel.

At least 19 recordsLinked to original sources

arXiv2026

Numerical approximations of population size distributions for multi-type branching processes

Continuous-time multi-type branching processes are fundamental models for expanding and migrating populations with cancer evolution being a prototypical example. Inferring model parameters, like mutation and growth rates, from time-series count data requires efficient computation of population size distributions. Existing methods are mainly based on large-time or large-number asymptotics, which rely on either restricted initial conditions or simplified interactions between cell types. Here, we introduce two numerical approximations of population size distributions for multi-type branching processes on directed graphs with arbitrary initialization. The first approach combines a saddle-point approximation with numerical integration of the probability generating function. We characterize admissibility and establish conditions for saddle-point existence and uniqueness. For directed acyclic graphs, the second approach provides a large-time small-mutation-rate alternative based on closed-form approximate Laplace transforms and efficient numerical inversion. We benchmark the accuracy and speed of both solutions in simulations, showing substantial improvement over the state-of-the-art large-number approximation and orders of magnitude speedup over Gillespie's stochastic simulation algorithm at matching accuracy. We apply our methods to analyze the relapse dynamics of an acute myeloid leukemia patient, where rapid parameter scans over a six-type patient-specific mutation tree quantify how unobserved remission burden and treatment-altered fitness can explain relapse. Our methods provide computational building blocks for future likelihood-based inference in cancer evolution and other expanding populations.

q-bio.PE↗

arXiv2025

HIVMedQA: Benchmarking large language models for HIV medical decision support

Large language models (LLMs) are emerging as valuable tools to support clinicians in routine decision-making. HIV management is a compelling use case due to its complexity, including diverse treatment options, comorbidities, and adherence challenges. However, integrating LLMs into clinical practice raises concerns about accuracy, potential harm, and clinician acceptance. Despite their promise, AI applications in HIV care remain underexplored, and LLM benchmarking studies are scarce. This study evaluates the current capabilities of LLMs in HIV management, highlighting their strengths and limitations. We introduce HIVMedQA, a benchmark designed to assess open-ended medical question answering in HIV care. The dataset consists of curated, clinically relevant questions developed with input from an infectious disease physician. We evaluated seven general-purpose and three medically specialized LLMs, applying prompt engineering to enhance performance. Our evaluation framework incorporates both lexical similarity and an LLM-as-a-judge approach, extended to better reflect clinical relevance. We assessed performance across key dimensions: question comprehension, reasoning, knowledge recall, bias, potential harm, and factual accuracy. Results show that Gemini 2.5 Pro consistently outperformed other models across most dimensions. Notably, two of the top three models were proprietary. Performance declined as question complexity increased. Medically fine-tuned models did not always outperform general-purpose ones, and larger model size was not a reliable predictor of performance. Reasoning and comprehension were more challenging than factual recall, and cognitive biases such as recency and status quo were observed. These findings underscore the need for targeted development and evaluation to ensure safe, effective LLM integration in clinical care.

cs.CL↗

arXiv2025

Practical and scalable simulations of non-Markovian stochastic processes and temporal networks with individual node properties

Discrete stochastic processes are prevalent in natural systems, with applications in physics, biochemistry, epidemiology, sociology, and finance. While analytic solutions often cannot be derived, existing simulation frameworks can generate stochastic trajectories compatible with the dynamical laws underlying the random phenomena. Still, most simulation algorithms assume the system dynamics are memoryless (Markovian assumption), under which assumption, future occurrences only depend on the system's present state. This enables efficient and exact simulation via the Gillespie algorithm. However, many real-world systems are inherently non-Markovian and exhibit memory effects. Such systems are difficult to study analytically, and current numerical methods are often computationally expensive or limited by strong simplifying assumptions that conflict with empirical data. To address these limitations, we introduce the Rejection-based Gillespie algorithm for non-Markovian Reactions (REGIR), a general and scalable framework for simulating non-Markovian stochastic systems with arbitrary inter-event time distributions. REGIR provides user-defined accuracy while preserving the same asymptotic computational complexity as the classical Gillespie algorithm. We derive a lower bound on REGIR's approximation accuracy and demonstrate its capabilities across three representative classes of non-Markovian systems: (1) reaction channels with delays, (2) stochastic processes driven by individual reactant properties, and (3) temporal networks governed by node activity. In all cases, REGIR accurately captures memory-dependent dynamics and outperforms existing approaches in terms of flexibility and efficiency.

q-bio.QM↗

arXiv2024

On the Properties and Estimation of Pointwise Mutual Information Profiles

The pointwise mutual information profile, or simply profile, is the distribution of pointwise mutual information for a given pair of random variables. One of its important properties is that its expected value is precisely the mutual information between these random variables. In this paper, we analytically describe the profiles of multivariate normal distributions and introduce a novel family of distributions, Bend and Mix Models, for which the profile can be accurately estimated using Monte Carlo methods. We then show how Bend and Mix Models can be used to study the limitations of existing mutual information estimators, investigate the behavior of neural critics used in variational estimators, and understand the effect of experimental outliers on mutual information estimation. Finally, we show how Bend and Mix Models can be used to obtain model-based Bayesian estimates of mutual information, suitable for problems with available domain expertise in which uncertainty quantification is necessary.

stat.ML↗

arXiv2023

Fair Clustering: A Causal Perspective

Clustering algorithms may unintentionally propagate or intensify existing disparities, leading to unfair representations or biased decision-making. Current fair clustering methods rely on notions of fairness that do not capture any information on the underlying causal mechanisms. We show that optimising for non-causal fairness notions can paradoxically induce direct discriminatory effects from a causal standpoint. We present a clustering approach that incorporates causal fairness metrics to provide a more nuanced approach to fairness in unsupervised learning. Our approach enables the specification of the causal fairness metrics that should be minimised. We demonstrate the efficacy of our methodology using datasets known to harbour unfair biases.

stat.ML↗

arXiv2023

Beyond Normal: On the Evaluation of Mutual Information Estimators

Mutual information is a general statistical dependency measure which has found applications in representation learning, causality, domain generalization and computational biology. However, mutual information estimators are typically evaluated on simple families of probability distributions, namely multivariate normal distribution and selected distributions with one-dimensional random variables. In this paper, we show how to construct a diverse family of distributions with known ground-truth mutual information and propose a language-independent benchmarking platform for mutual information estimators. We discuss the general applicability and limitations of classical and neural estimators in settings involving high dimensions, sparse interactions, long-tailed distributions, and high mutual information. Finally, we provide guidelines for practitioners on how to select appropriate estimator adapted to the difficulty of problem considered and issues one needs to consider when applying an estimator to a new data set.

stat.ML↗

arXiv2023

Genomic reproducibility in the bioinformatics era

In biomedical research, validation of a new scientific discovery is tied to the reproducibility of its experimental results. However, in genomics, the definition and implementation of reproducibility still remain imprecise. Here, we argue that genomic reproducibility, defined as the ability of bioinformatics tools to maintain consistent genomics results across technical replicates, is key to generating scientific knowledge and enabling medical applications. We first discuss different concepts of reproducibility and then focus on reproducibility in the context of genomics, aiming to establish clear definitions of relevant terms. We then focus on the role of bioinformatics tools and their impact on genomic reproducibility and assess methods of evaluating bioinformatics tools in terms of genomic reproducibility. Lastly, we suggest best practices for enhancing genomic reproducibility, with an emphasis on assessing the performance of bioinformatics tools through rigorous testing across multiple technical replicates.

q-bio.GN↗

arXiv2022

High-Dimensional Inference in Bayesian Networks

Inference of the marginal probability distribution is defined as the calculation of the probability of a subset of the variables and is relevant for handling missing data and hidden variables. While inference of the marginal probability distribution is crucial for various problems in machine learning and statistics, its exact computation is generally not feasible for categorical variables in Bayesian networks due to the NP-hardness of this task. We develop a divide-and-conquer approach using the graphical properties of Bayesian networks to split the computation of the marginal probability distribution into sub-calculations of lower dimensionality, thus reducing the overall computational complexity. Exploiting this property, we present an efficient and scalable algorithm for calculating the marginal probability distribution for categorical variables. The novel method is compared against state-of-the-art approximate inference methods in a benchmarking study, where it displays superior performance. As an immediate application, we demonstrate how our method can be used to classify incomplete data against Bayesian networks and use this approach for identifying the cancer subtype of kidney cancer patient samples.

stat.ML↗

arXiv2021

Bayesian structure learning and sampling of Bayesian networks with the R package BiDAG

The R package BiDAG implements Markov chain Monte Carlo (MCMC) methods for structure learning and sampling of Bayesian networks. The package includes tools to search for a maximum a posteriori (MAP) graph and to sample graphs from the posterior distribution given the data. A new hybrid approach to structure learning enables inference in large graphs. In the first step, we define a reduced search space by means of the PC algorithm or based on prior knowledge. In the second step, an iterative order MCMC scheme proceeds to optimize within the restricted search space and estimate the MAP graph. Sampling from the posterior distribution is implemented using either order or partition MCMC. The models and algorithms can handle both discrete and continuous data. The BiDAG package also provides an implementation of MCMC schemes for structure learning and sampling of dynamic Bayesian networks.

stat.CO↗

arXiv2018

Bayesian reconstruction of HIV transmission trees from viral sequences and uncertain infection times

Genetic sequence data of pathogens are increasingly used to investigate transmission dynamics in both endemic diseases and disease outbreaks; such research can aid in development of appropriate interventions and in design of studies to evaluate them. Several methods have been proposed to infer transmission chains from sequence data; however, existing methods do not generally reliably reconstruct transmission trees because genetic sequence data or inferred phylogenetic trees from such data are insufficient for accurate inference regarding transmission chains. In this paper, we demonstrate the lack of a one-to-one relationship between phylogenies and transmission trees, and also show that information regarding infection times together with genetic sequences permit accurate reconstruction of transmission trees. We propose a Bayesian inference method for this purpose and demonstrate that precision of inference regarding these transmission trees depends on precision of the estimated times of infection. We also illustrate the use of these methods to study features of epidemic dynamics, such as the relationship between characteristics of nodes and average number of outbound edges or inbound edges-- signifying possible transmission events from and to nodes. We study the performance of the proposed method in simulation experiments and demonstrate its superiority in comparison to an alternative method. We apply them to a transmission cluster in San Diego and investigate the impact of biological, behavioral, and demographic factors.

q-bio.QM↗

arXiv2015

TiMEx: A Waiting Time Model for Mutually Exclusive Groups of Cancer Alterations

Despite recent technological advances in genomic sciences, our understanding of cancer progression and its driving genetic alterations remains incomplete. Here, we introduce TiMEx, a generative probabilistic model for detecting patterns of various degrees of mutual exclusivity across genetic alterations, which can indicate pathways involved in cancer progression. TiMEx explicitly accounts for the temporal interplay between the waiting times to alterations and the observation time. In simulation studies, we show that our model outperforms previous methods for detecting mutual exclusivity. On large-scale biological datasets, TiMEx identifies gene groups with strong functional biological relevance, while also proposing many new candidates for biological validation. TiMEx possesses several advantages over previous methods, including a novel generative probabilistic model of tumorigenesis, direct estimation of the probability of mutual exclusivity interaction, computational efficiency, as well as high sensitivity in detecting gene groups involving low-frequency alterations. R code is available at www.cbg.bsse.ethz.ch/software/TiMEx.

q-bio.MN↗

arXiv2015

Probabilistic modeling of occurring substitutions in PAR-CLIP data

Photoactivatable ribonucleoside-enhanced cross-linking and immunoprecipitation (PAR-CLIP) is an experimental method based on next-generation sequencing for identifying the RNA interaction sites of a given protein. The method deliberately inserts T-to-C substitutions at the RNA-protein interaction sites, which provides a second layer of evidence compared to other CLIP methods. However, the experiment includes several sources of noise which cause both low-frequency errors and spurious high-frequency alterations. Therefore, rigorous statistical analysis is required in order to separate true T-to-C base changes, following cross-linking, from noise. So far, most of the existing PAR-CLIP data analysis methods focus on discarding the low-frequency errors and rely on high-frequency substitutions to report binding sites, not taking into account the possibility of high-frequency false positive substitutions. Here, we introduce BMix, a new probabilistic method which explicitly accounts for the sources of noise in PAR- CLIP data and distinguishes cross-link induced T-to-C substitutions from low and high-frequency erroneous alterations. We demonstrate the superior speed and accuracy of our method compared to existing approaches on both simulated and real, publicly available human datasets. The model is implemented in the Matlab toolbox BMix, freely available at www.cbg.bsse.ethz.ch/software/BMix.

q-bio.QM↗

arXiv2013

Covering Pairs in Directed Acyclic Graphs

The Minimum Path Cover problem on directed acyclic graphs (DAGs) is a classical problem that provides a clear and simple mathematical formulation for several applications in different areas and that has an efficient algorithmic solution. In this paper, we study the computational complexity of two constrained variants of Minimum Path Cover motivated by the recent introduction of next-generation sequencing technologies in bioinformatics. The first problem (MinPCRP), given a DAG and a set of pairs of vertices, asks for a minimum cardinality set of paths "covering" all the vertices such that both vertices of each pair belong to the same path. For this problem, we show that, while it is NP-hard to compute if there exists a solution consisting of at most three paths, it is possible to decide in polynomial time whether a solution consisting of at most two paths exists. The second problem (MaxRPSP), given a DAG and a set of pairs of vertices, asks for a path containing the maximum number of the given pairs of vertices. We show its NP-hardness and also its W[1]-hardness when parametrized by the number of covered pairs. On the positive side, we give a fixed-parameter algorithm when the parameter is the maximum overlapping degree, a natural parameter in the bioinformatics applications of the problem.

cs.DS↗

arXiv2011

Evolutionary Games with Affine Fitness Functions: Applications to Cancer

We analyze the dynamics of evolutionary games in which fitness is defined as an affine function of the expected payoff and a constant contribution. The resulting inhomogeneous replicator equation has an homogeneous equivalent with modified payoffs. The affine terms also influence the stochastic dynamics of a two-strategy Moran model of a finite population. We then apply the affine fitness function in a model for tumor-normal cell interactions to determine which are the most successful tumor strategies. In order to analyze the dynamics of concurrent strategies within a tumor population, we extend the model to a three-strategy game involving distinct tumor cell types as well as normal cells. In this model, interaction with normal cells, in combination with an increased constant fitness, is the most effective way of establishing a population of tumor cells in normal tissue.

q-bio.PE↗

arXiv2011

Competition-colonization trade-off promotes coexistence of low-virulence viral strains

RNA viruses exist as genetically diverse populations displaying different phenotypes, including diverse degrees of virulence. The evolution of virulence in viral populations is, however, poorly understood. Based on the experimental observation of an RNA virus clone in cell culture diversifying into two subpopulations of different virulence, we study the dynamics of heterogeneous virus populations and the evolution of virulence. We introduce a competition-colonization trade-off into standard mathematical models of intra-host viral infection. Colonizers are fast spreading, virulent strains, whereas competitors are less virulent variants that are more successful within coinfected cells. We observe biphasic dynamics of the population: Early in the infection the population is dominated by colonizers, which later will be outcompeted by competitors. The simulations suggest the existence of a steady state where few low-virulence variants coexist. This equilibrium implies collective virulence attenuation in the population, in contrast to previous models predicting development of the population towards increased virulence. Nevertheless, the attenuation effect disappears if we include a highly simplified immune response in our models. Thus, the competition-colonization trade-off indicates a role for virulence in the modulation of the viral population diversity. The evolution of virulence is a dynamic feature of the population shaped by interactions between individuals and by the structure of the patchy habitat.

q-bio.PE↗

arXiv2010

The evolution of virulence in RNA viruses under a competition-colonization trade-off

RNA viruses exist in large intra-host populations which display great genotypic and phenotypic diversity. We analyze a model of viral competition between two different viral strains infecting a constantly replenished cell pool, in which we assume a trade-off between the virus' colonization skills (cell killing ability or virulence) and its local competition skills (replication performance within coinfected cells). We characterize the conditions that allow for viral spread by means of the basic reproductive number and show that a local coexistence equilibrium exists, which is asymptotically stable. At this equilibrium, the less virulent competitor has a reproductive advantage over the more virulent colonizer. The equilibria at which one strain outcompetes the other one are unstable, i.e., a second viral strain is always able to permanently invade. One generalization of the model is to consider multiple viral strains, each one displaying a different virulence. However, to account for the large phenotypic diversity in viral populations, we consider a continuous spectrum of virulences and present a continuum limit of this multiple viral strains model that describes the time evolution of an initial continuous distribution of virulence. We provide a proof of the existence of solutions of the model's equations and present numerical approximations of solutions for different initial distributions. Our simulations suggest that initial continuous distributions of virulence evolve towards a stationary distribution that is extremely skewed in favor of competitors. Consequently, collective virulence attenuation takes place. This finding may contribute to understanding the phenomenon of virulence attenuation, which has been reported in previous experimental studies.

q-bio.PE↗

arXiv2009

Waiting time models of cancer progression

Cancer progression is an evolutionary process that is driven by mutation and selection in a population of tumor cells. We discuss mathematical models of cancer progression, starting from traditional multistage theory. Each stage is associated with the occurrence of genetic alterations and their fixation in the population. We describe the accumulation of mutations using conjunctive Bayesian networks, an exponential family of waiting time models in which the occurrence of mutations is constrained to a partial temporal order. Two opposing limit cases arise if mutations either follow a linear order or occur independently. We derive exact analytical expressions for the waiting time until a specific number of mutations have accumulated in these limit cases as well as for the general conjunctive Bayesian network. Finally, we analyze a stochastic population genetics model that explicitly accounts for mutation and selection. In this model, waves of clonal expansions sweep through the population at equidistant intervals. We present an approximate analytical expression for the waiting time in this model and compare it to the results obtained for the conjunctive Bayesian networks.

q-bio.PE↗

arXiv2009

Multiple Sequence Alignment System for Pyrosequencing Reads

Pyrosequencing is among the emerging sequencing techniques, capable of generating upto 100,000 overlapping reads in a single run. This technique is much faster and cheaper than the existing state of the art sequencing technique such as Sanger. However, the reads generated by pyrosequencing are short in size and contain numerous errors. Furthermore, each read has a specific position in the reference genome. In order to use these reads for any subsequent analysis, the reads must be aligned . Existing multiple sequence alignment methods cannot be used as they do not take into account the specific positions of the sequences with respect to the genome, and are highly inefficient for large number of sequences. Therefore, the common practice has been to use either simple pairwise alignment despite its poor accuracy for error prone pyroreads, or use computationally expensive techniques based on sequential gap propagation. In this paper, we develop a computationally efficient method based on domain decomposition, referred to as pyro-align, to align such large number of reads. The proposed alignment algorithm accurately aligns the erroneous reads in a short period of time, which is orders of magnitude faster than any existing method. The accuracy of the alignment is confirmed from the consensus obtained from the multiple alignments.

q-bio.GN↗