Biology subjects

Matijevic, D.

Publications and source records attributed to Matijevic, D..

5 recordsLinked to original sources

bioRxiv2026

An explanatory benchmark of spatial domain detection reveals key drivers of method performance

The spatial organization of cells within tissues is critical for understanding biological function and disease, and spatial transcriptomics enables genome-wide mapping of this organization. Numerous computational methods aim to identify spatial domains, yet their performance is often evaluated on limited datasets, leading to conflicting conclusions. Here, we present an explanatory benchmark of 26 spatial domain detection methods across 63 tissue sections from six spatial transcriptomics technologies, supplemented by over 1,000 semi-synthetic datasets that systematically vary resolution, gene panel size, and tissue architecture. By jointly analyzing real and semi-synthetic data across this broad parameter space, our benchmark uncovers systematic performance differences and sources of variability that are obscured in standard evaluations. Although most spatial methods outperform non-spatial baselines, their performance depends strongly on data resolution and cellular heterogeneity. To enable systematic analysis beyond individual methods, we introduce a modular, plug-and-play benchmarking framework that facilitates method refinement and component exchange. Using this framework, an ablation study of neural network-based approaches shows that the choice of preprocessing and clustering often has a larger impact on performance than ar-chitectural novelty alone. Together, these results provide a principled foundation for informed method selection and offer guidance for the development of robust and scalable spatial domain detection tools as spatial transcriptomics technologies continue to advance.

bioinformatics↗

bioRxiv2025

Optimal marker genes for c-separated cell types

The identification of cell types in single-cell RNA-seq studies relies on the distinct expression signature of marker genes. A small set of target genes is also needed to design probes for targeted spatial transcriptomics experiments and to target proteins in single-cell spatial proteomics or for cell sorting. While traditional approaches have relied on testing one gene at a time for differential expression between a given cell type and the rest, more recent methods have highlighted the benefits of a joint selection of markers that together distinguish all pairs of cell types simultaneously. These combinatorial methods differ mostly in the notion of discrimination between cell types, using Euclidean or Manhattan distance in the dimensions of the selected marker genes, or the difference in the fraction of cells expressing a given marker gene. The resulting combinatorial optimization problems then seeks to identify a small set of genes that yield discrimination above a given threshold between all pairs of cell types. However, existing methods either consider all pairs of individual cells which becomes intractable even for medium-sized datasets, or ignore intra-cell type expression variation entirely by collapsing all cells of a given type to a single representative. Here we address these limitations and propose to find a small set of genes such that cell types are c-separated in the selected dimensions, a notion introduced previously in learning a mixture of Gaussians. To this end, we formulate a linear program that naturally takes into account expression variation within cell types without including each pair of individual cells in the model, leading to a highly stable set of marker genes that allow to accurately discriminate between cell types and that can be computed to optimality efficiently.

bioinformatics↗

bioRxiv2023

Boosting N-terminally anchored yeast surface display via structural insights into S. cerevisiae Pir proteins

Surface display co-opts yeasts innate ability to embellish its cell wall with mannoproteins, thus converting the yeasts outer surface into a growing and self-sustaining catalyst. However, the efficient toolbox for converting the enzyme of interest into its surface-displayed isoform is currently lacking, especially if the isoform needs to be anchored to the cell wall near the isoforms N-terminus. Aiming to advance such N-terminally anchored surface display, we employed in silico and machine-learning strategies to study the 3D structure, function, genomic organisation, and evolution of the Pir protein family, whose members evolved to covalently attach themselves near their N-terminus to the {beta}-1,3-glucan of the cell wall. Through the newly-gained insights, we rationally engineered 14 S. cerevisiae Hsp150 (Pir2)-based fusion proteins. We quantified their performance, uncovering guidelines for efficient yeast surface display while developing a construct that promoted a 2.5-fold more efficient display than the full-length Hsp150 and a Pir-tag, i.e., a peptide spanning only 4.5 kDa but promoting as efficient surface display as the full-length Hsp150. These constructs fortify the existing surface display toolbox, allowing for a prompt and routine refitting of any protein into its N-terminally anchored isoform. Graphical abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=137 SRC="FIGDIR/small/538238v1_ufig1.gif" ALT="Figure 1"> View larger version (60K): org.highwire.dtl.DTLVardef@909e7aorg.highwire.dtl.DTLVardef@9480ceorg.highwire.dtl.DTLVardef@1945bdcorg.highwire.dtl.DTLVardef@11ae505_HPS_FORMAT_FIGEXP M_FIG C_FIG

molecular biology↗

bioRxiv2021

Metric Multidimensional Scaling for Large Single-Cell Data Sets using Neural Networks

Metric multidimensional scaling is one of the classical methods for embedding data into low-dimensional Euclidean space. It creates the low-dimensional embedding by approximately preserving the pairwise distances between the input points. However, current state-of-the-art approaches only scale to a few thousand data points. For larger data sets such as those occurring in single-cell RNA sequencing experiments, the running time becomes prohibitively large and thus alternative methods such as PCA are widely used instead. Here, we propose a neural network based approach for solving the metric multidimensional scaling problem that is orders of magnitude faster than previous state-of-the-art approaches, and hence scales to data sets with up to a few million cells. At the same time, it provides a non-linear mapping between high- and low-dimensional space that can place previously unseen cells in the same embedding.

bioinformatics↗

bioRxiv2019

Dynamic pseudo-time warping of complex single-cell trajectories

Single-cell RNA sequencing enables the construction of trajectories describing the dynamic changes in gene expression underlying biological processes such as cell differentiation and development. The comparison of single-cell trajectories under two distinct conditions can illuminate the differences and similarities between the two and can thus be a powerful tool. Recently developed methods for the comparison of trajectories rely on the concept of dynamic time warping (dtw), which was originally proposed for the comparison of two time series. Consequently, these methods are restricted to simple, linear trajectories. Here, we adopt and theoretically link arboreal matchings to dtw and propose an algorithm to compare complex trajectories that more realistically contain branching points that divert cells into different fates. We implement a suite of exact and heuristic algorithms suitable for the comparison of trajectories of different characteristics in our tool Trajan. Trajan automatically pairs similar biological processes between conditions and aligns them in a globally consistent manner. In an alignment of singlecell trajectories describing human muscle differentiation and myogenic reprogramming, Trajan identifies and aligns the core paths without prior information. From Trajans alignment, we are able to reproduce recently reported barriers to reprogramming. In a perturbation experiment, we demonstrate the benefits in terms of robustness and accuracy of our model which compares entire trajectories at once, as opposed to a pairwise application of dtw. Trajan is available at https://github.com/canzarlab/Trajan.

bioinformatics↗