Biology subjects

Kobak, D.

Publications and source records attributed to Kobak, D..

17 recordsLinked to original sources

bioRxiv2025

Principled PCA separates signal from noise in omics count data

Principal component analysis (PCA) is indispensable for processing high-throughput omics datasets, as it can extract meaningful biological variability while minimizing the influence of noise. However, the suitability of PCA is contingent on appropriate normalization and transformation of count data, and accurate selection of the number of principal components; improper choices can result in the loss of biological information or corruption of the signal due to excessive noise. Typical approaches to these challenges rely on heuristics that lack theoretical foundations. In this work, we present Biwhitened PCA (BiPCA), a theoretically grounded framework for rank estimation and data denoising across a wide range of omics modalities. BiPCA overcomes a fundamental difficulty with handling count noise in omics data by adaptively rescaling the rows and columns - a rigorous procedure that standardizes the noise variances across both dimensions. Through simulations and analysis of over 100 datasets spanning seven omics modalities, we demonstrate that BiPCA reliably recovers the data rank and enhances the biological interpretability of count data. In particular, BiPCA enhances marker gene expression, preserves cell neighborhoods, and mitigates batch effects. Our results establish BiPCA as a robust and versatile framework for high-throughput count data analysis.

bioinformatics↗

bioRxiv2024

Visualizing single-cell data with the neighbor embedding spectrum

The two-dimensional embedding methods t-SNE and UMAP are ubiquitously used for visualizing single-cell data. Recent theoretical research in machine learning has shown that, despite their very different formulation and implementation, t-SNE and UMAP are closely connected, and a single parameter suffices to interpolate between them. This leads to a whole spectrum of visualization methods that focus on different aspects of the data. Along the spectrum, this focus changes from representing local structures to representing continuous ones. In single-cell context, this leads to a trade-off between highlighting rare cell types or continuous variation, such as developmental trajectories. Visualizing the entire spectrum as an animation can provide a more nuanced understanding of the high-dimensional dataset than individual visualizations with either t-SNE or UMAP.

bioinformatics↗

bioRxiv2024

The art of seeing the elephant in the room: 2D embeddings of single-cell data do make sense

A recent paper in PLOS Computational Biology (Chari and Pachter, 2023) claimed that t-SNE and UMAP embeddings of single-cell datasets fail to capture true biological structure. The authors argued that such embeddings are as arbitrary and as misleading as forcing the data into an elephant shape. Here we show that this conclusion was based on inadequate and limited metrics of embedding quality. More appropriate metrics quantifying neighborhood and class preservation reveal the elephant in the room: while t-SNE and UMAP embeddings of single-cell data do not preserve high-dimensional distances, they can nevertheless provide biologically relevant information.

genomics↗

bioRxiv2023

Compound models and Pearson residuals for normalization of single-cell RNA-seq data without UMIs

Recent work employed Pearson residuals from Poisson or negative binomial models to normalize UMI data. To extend this approach to non-UMI data, we model the additional amplification step with a compound distribution: we assume that sequenced RNA molecules follow a negative binomial distribution, and are then replicated following an amplification distribution. We show how this model leads to compound Pearson residuals, which yield meaningful gene selection and embeddings of Smart-seq2 datasets. Further, we suggest that amplification distributions across several sequencing protocols can be described by a broken power law. The resulting compound model captures previously unexplained overdispersion and zero-inflation patterns in non-UMI data.

bioinformatics↗

bioRxiv2023

The landscape of biomedical research

The number of publications in biomedicine and life sciences has rapidly grown over the last decades, with over 1.5 million papers now being published every year. This makes it difficult to keep track of new scientific works and to have an overview of the evolution of the field as a whole. Here we present a 2D map of the entire corpus of biomedical literature, and argue that it provides a unique and useful overview of the life sciences research. We based our atlas on the abstract texts of 21 million English articles from the PubMed database. To embed the abstracts into 2D, we used the large language model PubMedBERT, combined with t-SNE tailored to handle samples of our size. We used our atlas to study the emergence of the Covid-19 literature, the evolution of the neuroscience discipline, the uptake of machine learning, the distribution of gender imbalance in academic authorship, and the distribution of retracted paper mill articles. Furthermore, we present an interactive web version of our atlas that allows easy exploration and will enable further insights and facilitate future research.

scientific communication and education↗

bioRxiv2023

Combined statistical-mechanistic modeling links ionchannel genes to physiology of cortical neuron types

Neural cell types have classically been characterized by their anatomy and electrophysiology. More recently, single-cell transcriptomics has enabled an increasingly fine genetically defined taxonomy of cortical cell types, but the link between the gene expression of individual cell types and their physiological and anatomical properties remains poorly understood. Here, we develop a hybrid modeling approach to bridge this gap. Our approach combines statistical and mechanistic models to predict cells electrophysiological activity from their gene expression pattern. To this end, we fit biophysical Hodgkin-Huxley-based models for a wide variety of cortical cell types using simulation-based inference, while overcoming the challenge posed by the mismatch between the mathematical model and the data. Using multimodal Patch-seq data, we link the estimated model parameters to gene expression using an interpretable sparse linear regression model. Our approach recovers specific ion channel gene expressions as predictive of biophysical model parameters including ion channel densities, directly implicating their mechanistic role in determining neural firing.

neuroscience↗

bioRxiv2020

Analytic Pearson residuals for normalization of single-cell RNA-seq UMI data

BackgroundStandard preprocessing of single-cell RNA-seq UMI data includes normalization by sequencing depth to remove this technical variability, and nonlinear transformation to stabilize the variance across genes with different expression levels. Instead, two recent papers propose to use statistical count models for these tasks: Hafemeister & Satija [1] recommend using Pearson residuals from negative binomial regression, while Townes et al. [2] recommend fitting a generalized PCA model. Here, we investigate the connection between these approaches theoretically and empirically, and compare their effects on downstream processing. ResultsWe show that the model of Hafemeister and Satija produces noisy parameter estimates because it is overspecified, which is why the original paper employs post-hoc smoothing. When specified more parsimoniously, it has a simple analytic solution equivalent to the rank-one Poisson GLM-PCA of Townes et al. Further, our analysis indicates that per-gene overdispersion estimates in Hafemeister and Satija are biased, and that the data are in fact consistent with the overdispersion parameter being independent of gene expression. We then use negative control data without biological variability to estimate the technical overdispersion of UMI counts, and find that across several different experimental protocols, the data are close to Poisson and suggest very moderate overdispersion. Finally, we perform a benchmark to compare the performance of Pearson residuals, variance-stabilizing transformations, and GLM-PCA on scRNA-seq datasets with known ground truth. ConclusionsWe demonstrate that analytic Pearson residuals strongly outperform other methods for identifying biologically variable genes, and capture more of the biologically meaningful variation when used for dimensionality reduction.

bioinformatics↗

bioRxiv2020

Distinct organization of two cortico-cortical feedback pathways

Neocortical feedback is critical for processes like attention, prediction, and learning. A mechanistic understanding of its function requires deciphering its cell-type wiring logic. Recent studies revealed a disinhibitory circuit between motor and sensory areas in mice, where feedback preferentially targets vasointestinal peptide-expressing interneurons, in addition to pyramidal cells. It is unknown whether this circuit motif is a general cortico-cortical feedback organizing principle. Combining multiple simultaneous whole-cell recordings with optogenetics we found that in contrast to this wiring rule, feedback between the hierarchically organized visual areas (lateral-medial to V1) preferentially activated somatostatin-expressing interneurons. Functionally, both feedback circuits temporally sharpened feed-forward excitation by eliciting a transient increase followed by a prolonged decrease in pyramidal firing rate under sustained feed-forward input. However, under feed-forward transient input, the motor-sensory feedback facilitated pyramidal cell bursting while visual feedback increased spike time precision. Our findings argue for multiple feedback motifs implementing different dynamic non-linear operations.

neuroscience↗

bioRxiv2020

Phenotypic variation within and across transcriptomic cell types in mouse motor cortex

Cortical neurons exhibit astounding diversity in gene expression as well as in morphological and electrophysiological properties. Most existing neural taxonomies are based on either transcriptomic or morpho-electric criteria, as it has been technically challenging to study both aspects of neuronal diversity in the same set of cells. Here we used Patch-seq to combine patch-clamp recording, biocytin staining, and single-cell RNA sequencing of over 1300 neurons in adult mouse motor cortex, providing a comprehensive morpho-electric annotation of almost all transcriptomically defined neural cell types. We found that, although broad families of transcriptomic types (Vip, Pvalb, Sst, etc.) had distinct and essentially non-overlapping morpho-electric phenotypes, individual transcriptomic types within the same family were not well-separated in the morpho-electric space. Instead, there was a continuum of variability in morphology and electrophysiology, with neighbouring transcriptomic cell types showing similar morpho-electric features, often without clear boundaries between them. Our results suggest that neural types in the neocortex do not always form discrete entities. Instead, neurons follow a hierarchy consisting of distinct non-overlapping branches at the level of families, but can form continuous and correlated transcriptomic and morpho-electrical landscapes within families.

neuroscience↗

bioRxiv2019

UMAP does not preserve global structure any better than t-SNE when using the same initialization

One of the most ubiquitous analysis tools employed in single-cell transcriptomics and cytometry is t-distributed stochastic neighbor embedding (t-SNE) [1], used to visualize individual cells as points on a 2D scatter plot such that similar cells are positioned close together. Recently, a related algorithm, called uniform manifold approximation and projection (UMAP) [2] has attracted substantial attention in the single-cell community. In Nature Biotechnology, Becht et al. [3] argued that UMAP is preferable to t-SNE because it better preserves the global structure of the data and is more consistent across runs. Here we show that this alleged superiority of UMAP can be entirely attributed to different choices of initialization in the implementations used by Becht et al.: t-SNE implementations by default used random initialization, while the UMAP implementation used a technique called Laplacian eigenmaps [4] to initialize the embedding. We show that UMAP with random initialization preserves global structure as poorly as t-SNE with random initialization, while t-SNE with informative initialization performs as well as UMAP with informative initialization. Hence, contrary to the claims of Becht et al., their experiments do not demonstrate any advantage of the UMAP algorithm per se, but rather warn against using random initialization.

bioinformatics↗

bioRxiv2019

A systematic evaluation of neural morphology representations forcell type discrimination

Quantitative analysis of neuronal morphologies usually begins with choosing a particular feature representation in order to make individual morphologies amenable to standard statistics tools and machine learning algorithms. Many different feature representations have been suggested in the literature, ranging from density maps to intersection profiles, but they have never been compared side by side. Here we performed a systematic comparison of various representations, measuring how well they were able to capture the difference between known morphological cell types. For our benchmarking effort, we used several curated data sets consisting of mouse retinal bipolar cells and cortical inhibitory neurons. We found that the best performing feature representations were two-dimensional density maps closely followed by morphometric statistics, which both continued to perform well even when neurons were only partially traced. The same representations performed well in an unsupervised setting, implying that they can be suitable for dimensionality reduction or clustering.

neuroscience↗

bioRxiv2019

Cell type composition and circuit organization of neocortical radial clones

SummaryExcitatory neurons arising from a common progenitor establish radially-oriented clonal units in the neocortex which have been proposed to serve as elementary information processing modules. To characterize the cell types and circuit diagram within these clonal units, we performed single-cell RNA-sequencing and multi-cell patch clamp recordings of neurons derived from Nestin-positive progenitors. We found that radial clones do not appear to be fate-restricted, but instead individual clones are composed of a random sampling of the transcriptomic cell types present in a particular cortical area. The effect of lineage on synaptic connectivity depends on the type of connection tested: pairs of clonally related neurons were more likely to be connected vertically, across cortical layers, but not laterally within the same layer, compared to unrelated pairs. We propose that integration of vertical input from related neurons with lateral input from unrelated neurons may represent a developmentally programmed motif for assembling neocortical circuits.

neuroscience↗

bioRxiv2018

Neocortical layer 4 in adult mouse differs in major cell types and circuit organization between primary sensory areas

Layer 4 (L4) of mammalian neocortex plays a crucial role in cortical information processing, yet a complete census of its cell types and connectivity remains elusive. Using whole-cell recordings with morphological recovery, we identified one major excitatory and seven inhibitory types of neurons in L4 of adult mouse visual cortex (V1). Nearly all excitatory neurons were pyramidal and almost all Somatostatin-positive (SOM+) neurons were Martinotti cells. In contrast, in somatosensory cortex (S1), excitatory cells were mostly stellate and SOM+ cells were non-Martinotti. These morphologically distinct SOM+ interneurons correspond to different transcriptomic cell types and are differentially integrated into the local circuit with only S1 cells receiving local excitatory input. Our results challenge the classical view of a canonical microcircuit repeated through the neocortex. Instead we propose that cell-type specific circuit motifs, such as the Martinotti/pyramidal pair, are optionally used across the cortex as building blocks to assemble cortical circuits.

neuroscience↗

bioRxiv2018

Motor skill learning leads to the increase of planning horizon

We investigated motor skill learning using a path tracking task, where human subjects had to track various curved paths at a constant speed while maintaining the cursor within the path width. Subjects accuracy increased with practice, even when tracking novel untrained paths. Using a "searchlight" paradigm, where only a short segment of the path ahead of the cursor was shown, we found that subjects with a higher tracking skill differed from the novice subjects in two respects. First, they had lower motor variability, in agreement with previous findings. Second, they took a longer section of the future path into account when performing the task, i.e. had a longer planning horizon. We estimate that between one third and one half of the performance increase was due to the increase in planning horizon. An optimal control model with a fixed horizon (receding horizon control) that increases with tracking skill quantitatively captured the subjects movement behaviour. These findings demonstrate that human subjects not only increase their motor acuity but also their planning horizon when acquiring a motor skill. New and NoteworthyWe show that when learning a motor skill humans are using information about the environment from an increasingly longer amount of the movement path ahead to improve performance. Crucial features of the behavioural performance can be captured by modelling the behavioural data with a receding horizon optimal control model.

neuroscience↗

bioRxiv2018

State-dependent geometry of population activity in rat auditory cortex

The accuracy of the sensory neural code depends on the relationship between variability in population activity caused by the stimuli versus variability caused by noise. Whereas a large body of work has addressed this issue at a theoretical level, there have been few systematic empirical investigations using large-scale simultaneous recordings. We studied the geometry of the population code in the rat auditory cortex using sound stimuli varying across two orthogonal dimensions: the difference in sound level between the two ears, and the average level across the ears. We quantified the structure of the code across recordings spanning a wide range of brain states along the activation/inactivation continuum. We found that the nature of the code, both in terms of the stimulus and the noise subspaces, depends strongly on the degree of cortical activation. During the inactive (or synchronized) state, the code is characterized by a prevalence of contra-lateral tuning, a preference for loud sounds, gain-modulated tuning curves and coherent global noise fluctuations overlapping with the signal subspace. In contrast, in the most active (or desynchronized) recordings, there is an approximately symmetric preference between contra-, ipsi-, quiet and loud sounds, no systematic changes in the gain of lateralization tuning curves by sound intensity, and noise fluctuations span dimensions orthogonal to those evoked by the stimuli. Logistic regression could predict the sound lateralization with accuracy similar to the behaving rats, but only during cortical activation. Furthermore, also exclusively during the active state, the lateralization of sounds of different overall level could be decoded with the same accuracy, even by the same decoder, due to an orthogonal representation of level-differences and overall level across the population. Decoding accuracy from surrogate datasets with independent noise across neurons still displays strong state-dependence, suggesting that tuning curve reorganization plays a critical role in the overall effect of brain state on the population code. Our results provide an empirical foundation for the geometry and state dependence of cortical population codes.

neuroscience↗

bioRxiv2018

The art of using t-SNE for single-cell transcriptomics

Single-cell transcriptomics yields ever growing data sets containing RNA expression levels for thousands of genes from up to millions of cells. Common data analysis pipelines include a dimensionality reduction step for visualising the data in two dimensions, most frequently performed using t-distributed stochastic neighbour embedding (t-SNE). It excels at revealing local structure in high-dimensional data, but naive applications often suffer from severe shortcomings, e.g. the global structure of the data is not represented accurately. Here we describe how to circumvent such pitfalls, and develop a protocol for creating more faithful t-SNE visualisations. It includes PCA initialisation, a high learning rate, and multi-scale similarity kernels; for very large data sets, we additionally use exaggeration and downsampling-based initialisation. We use published single-cell RNA-seq data sets to demonstrate that this protocol yields superior results compared to the naive application of t-SNE.

bioinformatics↗

bioRxiv2018

Sparse reduced-rank regression for exploratory visualization of single cell patch-seq recordings

In genomics, transcriptomics, and related biological fields (collectively known as omics), it is common to work with n << p datasets with the dimensionality much larger than the sample size. In recent years, combinations of experimental techniques began to yield multiple sets of features for the same set of biological replicates. One example is Patch-seq, a method combining single-cell RNA sequencing with electrophysiological recordings from the same cells. Here we present a framework based on sparse reduced-rank regression for obtaining an interpretable visualization of the relationship between the transcriptomic and the electrophysiological data. We use an elastic net regularization penalty that yields sparse solutions and allows for an efficient computational implementation. Using several publicly available Patch-seq datasets, we show that sparse reduced-rank regression outperforms both sparse full-rank regression and non-sparse reduced-rank regression in terms of predictive performance, and can outperform existing methods for sparse partial least squares and sparse canonical correlation analysis in terms of out-of-sample correlations. We introduce a bibiplot visualization in order to display the dominant factors determining the relationship between transcriptomic and electrophysiological properties of neurons. We believe that sparse reduced-rank regression can provide a valuable tool for the exploration and visualization of paired multivariate datasets, including Patch-seq.

bioinformatics↗