Biology subjects

Pfaffelhuber, P.

Publications and source records attributed to Pfaffelhuber, P..

8 recordsLinked to original sources

bioRxiv2026

In which cases does it make sense to classify an individual into one population and what are the alternatives?

During biogeographical ancestry (BGA) analysis, an unknown individual for whom ancestry informative markers (AIMs) were analyzed is often assigned to one of a few reference populations. For some individuals, such an assignment is not meaningful: the two parents of the individual may come from different populations, or the individual may come from a population not included in the reference database. In this study, we examined classification methods in combination with 104 AIMs from the VISAGE Enhanced Tool to provide one of the following three possible outcomes: (i) in-reference (IR), with an assignment to one of the reference populations; (ii) admixed (AD), with parents from two in-reference populations; and (iii) out-of-reference (OR), where the individual's ancestry is not sufficiently represented by the reference populations. We therefore compared seven classifiers in combination with class-specific acceptance thresholds for the OR decision. The best-performing method according to our selection criterion was a naive Bayes classifier over pairs of populations, which we call the recent naive Bayes (RNB) classifier. Populations held out of the reference panel were correctly rejected in $84.9\%$ of cases. We further tested the method on $800$ individuals from seven additional groups not included in the reference panel. Only $29.4\%$ of individuals from four non-admixed populations were rejected, resembling joint ancestry with reference populations. In addition, we investigated the prediction accuracy in a newly collected dataset of 28 individuals with self-reported ancestry by analyzing AIMs using massively parallel sequencing.

genetics↗

bioRxiv2025

Enhancing Intra-Continental Biogeographical Ancestry Prediction Through a Machine Learning Marker Selection Method

While classifiers such as TabPFN (Hollmann et al., 2025) and SNIPPER (Phillips et al., 2007a) achieve strong intercontinental performance (Heinzel et al., 2025), their accuracy in classifying individuals within Europe remains low. One major factor contributing to this limitation is the set of genetic markers used for classification. Marker panels such as the VISAGE Enhanced Tool (Xavier et al., 2022) are commonly employed in forensic genetics because they contain ancestry-informative markers (AIMs) that distinguish very well between major continental populations. However, these panels are often not optimized for fine-scale differentiation within continents, where genetic variation is more subtle and population structure is rather continuous. We apply machine learning to select informative markers for intra-European classification, using data from Consortium et al. (2015). Compared with the VISAGE Enhanced Tool and allele frequency-based approaches (Phillips et al., 2007b; Kosoy et al., 2009; Nassir et al., 2009; Kidd et al., 2014; Phillips et al., 2014a), our marker sets achieve substantially higher accuracy within Europe: For four European populations, accuracy improves from 68.2% (VISAGE, 104 markers) to 73.7% (100 new markers) and 82.3% (200 new markers). For five populations, accuracy rises from 56.1% (VISAGE) to 64.5% (100 new markers). Our results show that tailored marker selection markedly improves intra-continental classification. While optimized here for Europe, the method can be applied to any region with sufficient training data.

genetics↗

bioRxiv2025

Advancing Biogeographical Ancestry Predictions Through Machine Learning

Tools like Snipper or the Admixture Model count as state-of-the-art methods in forensic science for biogeographical ancestry. However, they have not been systematically compared to classifiers widely used in other disciplines. Noting that genetic data have a tabular form, this study addresses this gap by benchmarking forensic classifiers against TabPFN, a cutting-edge, general-purpose machine learning classifier for tabular data. The comparison evaluates performance using metrics such as accuracy--the proportion of correct classifications--and ROC AUC. We examine classification tasks for individuals at both the intracontinental and continental levels, based on a published dataset for training and testing. Our results reveal significant performance differences between methods, with TabPFN consistently achieving the best results for accuracy, ROC AUC and log loss. E.g., for accuracy, TabPFN improves SNIPPER from 84% to 93% on a continental scale using eight populations, and from 43% to 48% for inter-European classification with ten populations.

genetics↗

bioRxiv2021

Neural Networks for self-adjusting Mutation Rate Estimation when the Recombination Rate is unknown

Estimating the mutation rate, or equivalently effective population size, is a common task in population genetics. If recombination is low or high, optimal linear estimation methods are known and well understood. For intermediate recombination rates, the calculation of optimal estimators is more challenging. As an alternative to model-based estimation, neural networks and other machine learning tools could help to develop good estimators in these involved scenarios. However, if no benchmark is available it is difficult to assess how well suited these tools are for different applications in population genetics. Here we investigate feedforward neural networks for the estimation of the mutation rate based on the site frequency spectrum and compare their performance with model-based estimators. For this we use the model-based estimators introduced by Fu, Futschik et al., and Watterson that minimize the variance or mean square error for no and free recombination. We find that neural networks reproduce these estimators if provided with the appropriate features and training sets. Remarkably, using the model-based estimators to adjust the weights of the training data, only one hidden layer is necessary to obtain a single estimator that performs almost as well as model-based estimators for low and high recombination rates, and at the same time provides a superior estimation method for intermediate recombination rates. We apply the method to simulated data based on the human chromosome 2 recombination map, highlighting its robustness in a realistic setting where local recombination rates vary and/or are unknown.

evolutionary biology↗

bioRxiv2020

Inference of recent admixture using genotype data

The inference of biogeographic ancestry (BGA) has become a focus of forensic genetics. Misinference of BGA can have profound unwanted consequences for investigations and society. We show that recent admixture can lead to misclassification and erroneous inference of ancestry proportions, using state of the art analysis tools with (i) simulations, (ii) 1000 genomes project data, and (iii) two individuals analyzed using the ForenSeq DNA Signature Prep Kit. Subsequently, we extend existing tools for estimation of individual ancestry (IA) by allowing for different IA in both parents, leading to estimates of parental individual ancestry (PIA), and a statistical test for recent admixture. Estimation of PIA outperforms IA in most scenarios of recent admixture. Furthermore, additional information about parental ancestry can be acquired with PIA that may guide casework. HighlightsO_LIWe improve statistical methods as used in STRUCTURE and ADMIXTURE for Biogeographical Ancestry (BGA) inference to account for recent admixture, i.e. different admixture of both parents. C_LIO_LIThe resulting recent admixture model has a higher accuracy in estimating individual admixture in most cases. C_LIO_LIWe give a likelihood ratio test for recent admixture, which is both highly specific and sensitive for recent admixture. C_LIO_LIWe find evidence of recent admixture in the 1000 genomes dataset. C_LIO_LIThe self-report on recent admixture of two self-sequenced samples was only confirmed in one case. C_LI

genetics↗

bioRxiv2019

How to choose sets of ancestry informative markers: A supervised feature selection approach

Inference of the Biogeographical Ancestry (BGA) of a person or trace relies on three ingredients: (1) A reference database of DNA samples including BGA information; (2) a statistical clustering method; (3) a set of loci which segregate dependent on geographical location, i.e. a set of so-called Ancestry Informative Markers (AIMs). We used the theory of feature selection from statistical learning in order to obtain AIM-sets for BGA inference. Using simulations, we show that this learning procedure works in various cases, and outperforms ad hoc methods, based on statistics like FST or informativeness for the choice of AIMs. Applying our method to data from the 1000 genomes project (excluding Admixed Americans) we identified an AIMset of 17 SNPs, which partly overlaps with existing ones. For continental BGA, the AIMset outperforms existing AIMsets on the 1000 genomes dataset, and gives a vanishing misclassification error.

bioinformatics↗

bioRxiv2018

Genealogical distances under low levels of selection

For a panmictic population of constant size evolving under neutrality, Kingmans coalescent describes the genealogy of a population sample in equilibrium. However, for genealogical trees under selection, not even expectations for most basic quantities like height and length of the resulting random tree are known. Here, we give an analytic expression for the distribution of the total tree length of a sample of size n under low levels of selection in a two-alleles model. We can prove that trees are shorter than under neutrality under genic selection and if the beneficial mutant has dominance h < 1/2, but longer for h > 1/2. The difference to neutrality is [Formula] for genic selection with selection intensity and [Formula] for other modes of dominance.

evolutionary biology↗

bioRxiv2017

Limits of noise for autoregulated gene expression

Gene expression is influenced by extrinsic noise (involving a fluctuating environment of cellular processes) and intrinsic noise (referring to fluctuations within a cell under constant environment). We study the standard model of gene expression including an (in-)active gene, mRNA and protein. Gene expression is regulated in the sense that the protein feeds back and either represses (negative feedback) or enhances (positive feedback) its production at the stage of transcription. While it is well-known that negative (positive) feedback reduces (increases) intrinsic noise, we give a precise result on the resulting fluctuations in protein numbers. The technique we use is an extension of the Langevin approximation and is an application of a central limit theorem under stochastic averaging for Markov jump processes (Kang, Kurtz and Popovic, 2014). We find that (under our scaling and in equilibrium), negative feedback leads to a reduction in the Fano factor of at most 2, while the noise under positive feedback is potentially unbounded. The fit with simulations is very good and improves on known approximations.

systems biology↗