Biology subjects

Dekker, J.

Publications and source records attributed to Dekker, J..

At least 19 recordsLinked to original sources

bioRxiv2026

Robust V(D)J Recombination in the Absence of SA1/2-mediated Chromatin Loops

The cohesin ring complex organizes and mediates functional activity of the genome through chromatin loop extrusion1,2. Cohesin comprises a core SMC1-SMC3-RAD21 ring associated with multiple regulatory proteins, including SA1 and SA2 paralogs that bind mutually-exclusively to a RAD21 site3,4. Functional studies of cohesin complexes in vivo often employ inactivation of particular components2,3. However, inactivation of SA1 or SA2 in mice is limited by mouse embryonic lethality5-7 and, for dual SA1/2-inactivation by cycling cell lethality8-10. For these reasons, it remains unknown whether the requirement for SA1/2 in loop extrusion in vitro11,12 applies in vivo1,2, and whether these factors have overlapping or distinct functions in regulating this process3. Here, we employ G0/G1-arrested v-Abl pro-B lines, which are viable in the absence of cohesin-complex factors13,14, to dissect SA1 and SA2 functions in V(D)J recombination, a strictly G0/G1-specific process that harnesses loop extrusion to effect distinct Igh and Ig{kappa} locus recombination mechanisms15,16. Robust long-range Igh or Ig{kappa} V(D)J recombination continued unabated in the absence of SA1 or SA2. However, inactivation of both abrogated most positioned chromatin loops genome-wide and also abolished long-range Igh V(D)J recombination with a phenotype similar to RAD21 (cohesin) inactivation13. Yet, robust Ig{kappa} V(D)J recombination continued upon dual SA1/2-inactivation, which along with Hi-C findings, demonstrates ongoing loop-extrusion in the absence of SA1/2. Correspondingly, replacement of weak Igh V(D)J recombination signal sequences (RSSs) that target V(D)J recombination activity, with strong Ig{kappa} counterparts15 restored long-range Igh V(D)J recombination in SA1/2 inactivated cells. We conclude that SA1/2 together maintain loop extrusion impediments and genome-wide architecture in G0/G1-arrested cells; and, in their absence, loop extrusion proceeds unimpeded. Remarkably, strong RSSs allow Ig{kappa} to function normally in the absence of internal positioned loop structure across the 3.1-mega-base Ig{kappa} locus. We discuss implications of our findings for long-range gene regulation in G0/G1-arrested cells more broadly.

immunology↗

bioRxiv2026

Stepwise reorganization of chromosome conformation and nuclear organization during stem cell differentiation

Nuclear organization is a fundamental feature of cell identity and cell fate determination. Compared with differentiated cells, pluripotent stem cells exhibit markedly distinct nuclear architecture at multiple levels, including chromosome folding, lamina association, histone modification landscapes and nuclear body organization. Although individual aspects of this reorganization have been characterized, how and in what order these features are remodeled as cells commit to a lineage remains poorly understood. Here, we temporally map nuclear reorganization across five stages of an in vitro differentiation system of human embryonic stem cells through definitive endoderm and hepatoblast intermediates into hepatocyte-like cells. By integrating Hi-C, RNA-seq, ATAC-seq, ChIP-seq, and CUT&RUN data, we establish a genome-wide framework linking structural, epigenetic, and transcriptional changes across differentiation. Combined with immunofluorescence imaging, chromosome painting, and liquid chromatin Hi-C (LC-Hi-C) we show that nuclear reorganization proceeds in a stepwise and temporally ordered manner through three major transitions. In the first transition, chromosomes condense into defined territories concurrent with transient anchoring of centromere-proximal regions to the nuclear periphery. In a second transition, active and inactive chromatin segregate, leading to stronger compartmentalization. This coincides with deposition and peripheralization of H3K9me2-marked chromatin and morphological changes in nuclear speckles, while chromatin conformation at speckle-associated regions stabilizes. In a third transition, after lineage-specific genes are activated, chromatin interactions globally stabilize, establishing a more stable nuclear architecture on top of these earlier large-scale structural rearrangements. Together, our results define a stepwise framework for nuclear reorganization during human embryonic stem cell differentiation and reveal that the transition from pluripotency to a differentiated state proceeds through coordinated, temporally ordered structural events.

genomics↗

bioRxiv2026

Formation of an RNA-mediated nuclear compartment

Mammalian nuclei are spatially compartmentalized so that active and inactive segments of the genome occupy separate sub-nuclear neighborhoods. Compartments are often found in association with nuclear structures such as the nuclear lamina, nucleoli, and nuclear speckles (speckles), suggesting links between them. The molecular mechanisms by which compartments form remain largely unknown. Speckles are nuclear bodies that contain high concentrations of RNA splicing factors and associated chromatin has a high density of highly expressed genes. Combining liquid chromatin Hi-C to quantify chromatin interaction lifetimes genome-wide, immunofluorescence, fluorescence in situ hybridization, live cell imaging, and nascent transcript analysis, we have determined the biophysical and molecular basis of the speckle-associated chromosomal compartment. We find that genomic regions making up this compartment are stably glued together. Surprisingly, removal of speckles by rapid depletion of SON and SRRM2 that form the structural scaffold of these bodies shows that this stable association is not dependent on the speckle itself. Instead, we find that RNA molecules are the molecular glue that forms the biophysical basis of the speckle chromatin compartment. Based on observations that promoters and enhancers are also engaged in stable long-lived chromatin interactions that are independent of RNA, we propose a pathway for formation of the speckle chromosomal compartment: Initial stable clustering of promoters and enhancers is followed by production of (nascent) RNA. The exceptionally high density of GC-rich RNA emerging from speckle-associated loci forms a glue that holds these loci together and facilitates recruitment of speckle components. The result is a structurally stable nuclear compartment that facilitates efficient splicing.

genomics↗

bioRxiv2026

Cohesin promotes genomic stability by suppressing unequal sister chromatid exchange

The protein complex cohesin plays critical roles in genomic stability by tethering together sister chromatids at their pericentric regions and along their arms from S phase until anaphase. Cohesin-mediated pericentric cohesion prevents aneuploidy by ensuring bipolar attachment of sister kinetochores. Arm cohesion prevents loss of heterozygosity by biasing DNA repair via recombination between sister chromatids rather than between homologs. Here, we investigate in yeast whether cohesin also enhances genomic stability by suppressing unequal sister chromatid exchange (USCE) between repetitive sequences. In wild-type cells, the USCE rate between repeats 4kb apart (proximal) was 15X higher than repeats 68kb apart (distal). The USCE between distal repeats but not proximal repeats increased 4 to 7-fold in mutants with altered cohesin subunits or auxiliary factors. The level of increased distal USCE corresponded with reduced arm cohesion, reduced density of cohesion arm sites, and higher sister loci mobility. Our results suggest that high density of arm cohesion sites confines repair of DNA damage to local sequences. When the density of cohesion sites decreases, sister chromatid sequences are less confined, thereby enhancing distal repeat interactions and USCE. Another set of mutations disrupted both DNA replication and cohesin loading at the replication fork during S phase. Remarkably, distal USCE in these mutants increased approximately 100-fold and was 6-fold more likely than proximal USCE. This preferential hyperdistal USCE can be explained by an aberrant sister-chromatid structure that is normally prevented by proper coupling of cohesin function and replication.

genetics↗

bioRxiv2026

UnionLoops: a workflow for calling chromatin loops across related Hi-C datasets with improved specificity, precision, and sensitivity

Chromatin loop calling from chromatin interaction data often exhibits substantial variability across related samples. We present UnionLoops, a computational workflow for chromatin loop calling across multiple related samples. UnionLoops integrates information across datasets to determine positions and dataset-specificity of looping interactions. It constructs a unified candidate loop set, applies consistent filtering and aggregation, and evaluates loop support across samples. We demonstrate that UnionLoops increases sensitivity for detecting shared chromatin loops, reduces spurious sample-specific calls, and improves concordance with independent genomic features, including CTCF and cohesin occupancy. UnionLoops enables improved biological interpretation of chromatin loop organization and dynamics across related conditions.

genomics↗

bioRxiv2025

Dynamic changes in chromosome and nuclear architecture during maturation of normal and ALS C9orf72 motor neurons

We have investigated changes in chromosome conformation, nuclear organization, and transcription during differentiation and maturation of control and mutant motor neurons harboring hexanucleotide expansions in the C9orf72 gene that cause amyotrophic lateral sclerosis (ALS). Using an in vitro reprogramming, differentiation and neural maturation protocol, we obtained highly purified populations of post-mitotic motor neurons for both normal and diseased cells. As expected, as fibroblasts are reprogrammed into iPSCs, and as iPSCs differentiate into motor neurons, chromatin accessibility, chromosome conformation, and nuclear organization change along with large-scale alterations in transcriptional profiles. We find that the transcriptome changes extensively during the first three weeks of post-mitotic neuronal maturation, with thousands of genes changing expression, but then is relatively stable for the next three weeks. In contrast, chromosome conformation and nuclear organization continue to change over the entire 6-week maturation period: chromosome territoriality increases, long-range interactions along chromosomes decrease, compartmentalization strength increases, and centromeres and telomeres increasingly cluster. In motor neurons derived from ALS patients such changes in chromosome conformation were much reduced. Chromatin accessibility changes also showed delayed maturation. The transcriptome in these cells matured relatively normally but with notable changes in expression of genes involved in lipid, sterol and mitochondrial function. We conclude that neural maturation is associated with large scale post-mitotic changes in gene expression, chromosome conformation and nuclear organization, and that these processes are defective in motor neurons derived from ALS patients carrying C9orf72 hexanucleotide repeat expansions.

neuroscience↗

bioRxiv2025

esBAF and INO80C fine-tune subcompartments and differentially regulate enhancer-promoter interactions

The genome is compacted in the nucleus through a hierarchical chromatin organization, ranging from chromosome territories to compartments, topologically associating domains (TADs), and individual nucleosomes. Nucleosome remodeling complexes hydrolyze ATP to translocate DNA and thereby mobilize histone proteins. While nucleosome remodeling complexes have been extensively studied for their roles in regulating nucleosome positioning and accessibility, their contributions to higher-order chromatin architecture remain less well understood. Here, we investigate the roles of two key nucleosome remodelers, esBAF and INO80C, in shaping 3D genome organization in mouse embryonic stem cells. Using Hi-C, we find that loss of either remodeler has minimal effects on global compartment or TAD structures. In contrast, subcompartment organization is notably altered, suggesting that esBAF and INO80C contribute to finer-scale chromatin topology. To overcome the limited resolution of Hi-C for detecting regulatory loops, we employed promoter capture Micro-C (PCMC), which revealed that the loss of esBAF or INO80C alters a subset of promoter anchored looping interactions. Although these changes occur at distinct genomic loci for each remodeler, the affected sites are commonly enriched for bivalent chromatin regions bound by OCT4, SOX2, and NANOG (OSN), as well as BRG1 and INO80 themselves. Together, our findings reveal that esBAF and INO80C selectively influence subcompartment identity and enhancer-promoter communication at key regulatory loci, highlighting a previously underappreciated role for nucleosome remodelers in higher-order chromatin organization.

genomics↗

bioRxiv2025

Histone Acetylation Differentially Modulates CTCF-CTCF Loops and Intra-TAD Interactions

The cohesin complex structures the interphase genome by extruding loops and organizing topologically associating domains (TADs). While cohesin engages chromatin in context- dependent modes, the regulatory influence of chromatin state on these interactions remains unclear. Here, we show that histone hyperacetylation, induced by the histone deacetylase inhibitor trichostatin A (TSA), preferentially disrupts short-range interactions within TADs but spares CTCF-anchored loops, despite reduced cohesin occupancy at these sites. These findings point to two functionally distinct cohesin populations: a TSA- sensitive pool within TADs, likely representing extruding, non-topologically bound cohesin, and a TSA-resistant population at CTCF-CTCF anchors that maintains loops through topological entrapment. Using a semi-in vitro system with TEV-cleavable RAD21, we show that TSA-resistant cohesin at CTCF sites becomes TSA-sensitive after proteolytic cleavage that opens the cohesin ring, showing that it is the topological engagement with DNA that makes cohesin, and CTCF-CTCF loops, TSA-resistant. Notably, we also detect TSA-sensitive cohesin at CTCF sites, suggesting the presence of transient, non-encircling cohesin that either precedes conversion to the stable form or is halted by pre-existing encircling cohesin. Together, our results suggest that cohesin exists in distinct biochemical states: an extruding form found within TADs and at CTCF sites, that is sensitive to hyperacetylation, and a topologically bound form specifically at CTCF-CTCF loops that is insensitive. The former may allow dynamic changes in chromatin loops, while latter ensures robustness of CTCF-anchored loops in response to chromatin state changes.

genomics↗

bioRxiv2024

An integrated view of the structure and function of the human 4D nucleome

The dynamic three-dimensional (3D) organization of the human genome (the "4D Nucleome") is closely linked to genome function. Here, we integrate a wide variety of genomic data generated by the 4D Nucleome Project to provide a detailed view of human 3D genome organization in widely used embryonic stem cells (H1-hESCs) and immortalized fibroblasts (HFFc6). We provide extensive benchmarking of 3D genome mapping assays and integrate these diverse datasets to annotate spatial genomic features across scales. The data reveal a rich complexity of chromatin domains and their sub-nuclear positions, and over one hundred thousand structural loops and promoter-enhancer interactions. We developed 3D models of population-based and individual cell-to-cell variation in genome structure, establishing connections between chromosome folding, nuclear organization, chromatin looping, gene transcription, and DNA replication. We demonstrate the use of computational methods to predict genome folding from DNA sequence, uncovering potential effects of genetic variants on genome structure and function. Together, this comprehensive analysis contributes insights into human genome organization and enhances our understanding of connections between the regulation of genome function and 3D genome organization in general.

genomics↗

bioRxiv2024

Interphase chromosome conformation is specified by distinct folding programs inherited via mitotic chromosomes or through the cytoplasm

Identity-specific interphase chromosome conformation must be re-established each time a cell divides. To understand how interphase folding is inherited, we developed an experimental approach that physically segregates mediators of G1 folding that are intrinsic to mitotic chromosomes from cytoplasmic factors. Proteins essential for nuclear transport, RanGAP1 and Nup93, were degraded in pro-metaphase arrested DLD-1 cells to prevent the establishment of nucleo-cytoplasmic transport during mitotic exit and isolate the decondensing mitotic chromatin of G1 daughter cells from the cytoplasm. Using this approach, we discover a transient folding intermediate entirely driven by chromosome-intrinsic factors. In addition to conventional compartmental segregation, this chromosome-intrinsic folding program leads to prominent genome-scale microcompartmentalization of mitotically bookmarked and cell type-specific cis-regulatory elements. This microcompartment conformation is formed during telophase and subsequently modulated by a second folding program driven by factors inherited through the cytoplasm in G1. This nuclear import-dependent folding program includes cohesin and factors involved in transcription and RNA processing. The combined and inter-dependent action of chromosome-intrinsic and cytoplasmic inherited folding programs determines the interphase chromatin conformation as cells exit mitosis.

genomics↗

bioRxiv2024

Polymer physics models reveal structural folding features of single-molecule gene chromatin conformations

Here, we employ polymer physics models of chromatin to investigate the 3D folding of a 2Mb wide genomic region encompassing the human LTN1 gene, a crucial DNA locus involved in key cellular functions. Through extensive Molecular Dynamics simulations, we reconstruct in-silico the ensemble of single-molecule LTN1 3D structures, which we benchmark against recent in-situ Hi-C 2.0 data. The model-derived single molecules are then used to predict structural folding features at the single-cell level, providing testable predictions for super-resolution microscopy experiments.

molecular biology↗

bioRxiv2024

The non-canonical thioreductase TMX2 is essential for neuronal survival during embryonic brain development

Biallelic variants in thioredoxin-related transmembrane 2 protein (TMX2) can cause a brain malformation of cortical development (MCD), characterized by primary microcephaly, polymicrogyria and pachygyria by an unknown mechanism. To better understand and visualize how TMX2 loss disrupts brain development in vivo we investigated the function of TMX2, using the zebrafish embryo as a model system. We generated zebrafish deficient for TMX2 ortholog tmx2b, which during the first 2 days post fertilization (dpf) showed normal behavioral activity and brain developmental hallmarks. From 3 dpf onwards however, tmx2b mutants failed to exhibit locomotor activity, which was accompanied by cell death primarily in the brain, but not in other organs or in the spinal cord. Strikingly, cell death in tmx2b mutants occurs specifically in newborn neurons within a [~]1.5-hour timeframe, whereas neuronal progenitor and other glial cells are preserved, and could be suppressed by inhibiting neuronal activity. In vivo GCaMP6s calcium imaging showed a persistent [~]2-fold increase in calcium in neurons after the onset of cell death. This suggests that calcium homeostasis underlies the tmx2b mutant brain phenotype. Altogether, our results indicate that TMX2 is an evolutionary conserved, protective regulator essential specifically for newborn neurons to survive after their differentiation in the vertebrate embryonic brain.

developmental biology↗

bioRxiv2024

Rules of engagement for condensins and cohesins guide mitotic chromosome formation

During mitosis, interphase chromatin is rapidly converted into rod-shaped mitotic chromosomes. Using Hi-C, imaging, proteomics and polymer modeling, we determine how the activity and interplay between loop-extruding SMC motors accomplishes this dramatic transition. Our work reveals rules of engagement for SMC complexes that are critical for allowing cells to refold interphase chromatin into mitotic chromosomes. We find that condensin disassembles interphase chromatin loop organization by evicting or displacing extrusive cohesin. In contrast, condensin bypasses cohesive cohesins, thereby maintaining sister chromatid cohesion while separating the sisters. Studies of mitotic chromosomes formed by cohesin, condensin II and condensin I alone or in combination allow us to develop new models of mitotic chromosome conformation. In these models, loops are consecutive and not overlapping, implying that condensins do not freely pass one another but stall upon encountering each other. The dynamics of Hi-C interactions and chromosome morphology reveal that during prophase loops are extruded in vivo at [~]1-3 kb/sec by condensins as they form a disordered discontinuous helical scaffold within individual chromatids.

molecular biology↗

bioRxiv2024

mRNA initiation and termination are spatially coordinated

The expression of a precise mRNA transcriptome is crucial for establishing cell identity and function, with dozens of alternative isoforms produced for a single gene sequence. The regulation of mRNA isoform usage occurs by the coordination of co-transcriptional mRNA processing mechanisms across a gene. Decisions involved in mRNA initiation and termination underlie the largest extent of mRNA isoform diversity, but little is known about any relationships between decisions at both ends of mRNA molecules. Here, we systematically profile the joint usage of mRNA transcription start sites (TSSs) and polyadenylation sites (PASs) across tissues and species. Using both short and long read RNA-seq data, we observe that mRNAs preferentially using upstream TSSs also tend to use upstream PASs, and congruently, the usage of downstream sites is similarly paired. This observation suggests that mRNA 5 end choice may directly influence mRNA 3 ends. Our results suggest a novel "Positional Initiation-Termination Axis" (PITA), in which the usage of alternative terminal sites are coupled based on the order in which they appear in the genome. PITA isoforms are more likely to encode alternative protein domains and use conserved sites. PITA is strongly associated with the length of genomic features, such that PITA is enriched in longer genes with more area devoted to regions that regulate alternative 5 or 3 ends. Strikingly, we found that PITA genes are more likely than non-PITA genes to have multiple, overlapping chromatin structural domains related to pairing of ordinally coupled start and end sites. In turn, PITA coupling is also associated with fast RNA Polymerase II (RNAPII) trafficking across these long gene regions. Our findings indicate that a combination of spatial and kinetic mechanisms couple transcription initiation and mRNA 3 end decisions based on ordinal position to define the expression mRNA isoforms.

genomics↗

bioRxiv2023

Benchmarking the identification of a single degraded protein to explore optimal search strategies for ancient proteins

Palaeoproteomics is a rapidly evolving discipline, and practitioners are constantly developing novel strategies for the analyses and interpretations of complex, degraded protein mixtures. The community has also established standards of good practice to interrogate our data. However, there is a lack of a systematic exploration of how these affect the identification of peptides, post-translational modifications (PTMs), proteins and their significance (through the False Discovery Rate) and correctness. We systematically investigated the performance of a wide range of sequencing tools and search engines in a controlled system: the experimental degradation of the single purified bovine {beta}-lactoglobulin (BLG), heated at 95 {degrees}C and pH 7 for 0, 4 and 128 days. We target BLG since it is one of the most robust and ubiquitous proteins in the archaeological record. We tested different reference database choices, a targeted dairy protein one, and the whole bovine proteome and the three digestion options (tryptic-, semi-tryptic- and non-specific searches), in order to evaluate the effects of search space and the identification of peptides. We also explored alternative strategies, including open search that allows for the global identification of PTMs based upon wide precursor mass tolerance and de novo sequencing to boost sequence coverage. We analysed the samples using Mascot, MaxQuant, Metamorpheus, pFind, Fragpipe and DeNovoGUI (pepNovo+, DirecTag, Novor), benchmarked these tools and discuss the optimal strategy for the characterisation of ancient proteins. We also studied physicochemical properties of the BLG that correlate with bias in the identification coverage.

bioinformatics↗

bioRxiv2023

Mitotic chromosomes harbor cell type and species-specific structural features within a universal looping architecture

The architecture of mammalian mitotic chromosomes is considered to be universal across species and cell types. However, some studies suggest that features of mitotic chromosomes might be cell type or species specific. We previously reported that CTCF binding in human differentiated cell lines is lost in mitosis, whereas mouse embryonic stem cells (mESC) display prominent binding at a subset of CTCF sites in mitosis. Here, we perform parallel footprint ATAC-seq data analyses of mESCs and somatic mouse and human cells to further explore these differences. We then investigate roles of mitotically bound (bookmarked) CTCF in prometaphase chromosome organization by Hi-C. We do not find any remaining interphase structures such as TADs or CTCF loops at mitotically bookmarked CTCF sites in mESCs. This suggests that mitotic loop extruders condensin I and II are not blocked by bound CTCF, and thus that any remaining CTCF binding does not alter mitotic chromosome folding. Lastly, we compare mitotic Hi-C data generated in this study in mouse with publicly available data from human and chicken cell lines. We do not find any cell type specific differences; however, we do find a difference between species. The average genomic size of mitotic loops is much smaller in chicken (200-350 kb), compared to human (500-750 kb) and mouse (1-2 mb). Interestingly, we find that this difference in loop size is correlated with the average genomic length of the q-arm in these species, a finding we confirm by microscopy measurements of chromosome compaction. This suggests that the dimensions of mitotic chromosomes can be modulated through control of sizes of loops generated by condensins to facilitate species-appropriate shortening of chromosome arms.

genomics↗

bioRxiv2023

Enhancer-driven local 3D chromatin domain folding modulates transcription in human mammary tumor cells

The genome is organized in functional compartments and structural domains at the sub-megabase scale. How within these domains interactions between numerous cis-acting enhancers and promoters regulate transcription remains an open question. Here, we determined chromatin folding and composition over several hundred kb around estrogen responsive genes in human breast cancer cell lines following hormone stimulation. Modeling of 5C data at 1.8 kb resolution was combined with quantitative 3D analysis of multicolor FISH measurements at 100 nm resolution and integrated with ChIP-seq data on transcription factor binding and histone modifications. We found that rapid estradiol induction of the progesterone gene (PGR) expression occurs in the context of pre-existing, cell type specific chromosomal architectures encompassing the 90 kb PGR coding region and an enhancer-spiked 5 300 kb upstream genomic region. In response to estradiol, interactions between estrogen-receptor (ER) bound regulatory elements are re-enforced. While initial enhancer - gene contacts coincide with RNA Pol 2 binding and transcription initiation, sustained hormone stimulation promotes ER accumulation creating a regulatory hub stimulating transcript synthesis. In addition to implications for estrogen receptor signaling, we uncover that preestablished chromatin architectures efficiently regulate gene expression upon stimulation without the need for de-novo extensive rewiring of long-range chromatin interactions.

cell biology↗

bioRxiv2023

A comprehensive experimental comparison between federated and centralized learning

PurposeFederated learning is an upcoming machine learning paradigm which allows data from multiple sources to be used for training of classifiers without the data leaving the source it originally resides. This can be highly valuable for use cases such as medical research, where gathering data at a central location can be quite complicated due to privacy and legal concerns of the data. In such cases, federated learning has the potential to vastly speed up the research cycle. Although federated and central learning have been compared from a theoretical perspective, an extensive experimental comparison of performances and learning behavior still lacks. MethodsWe have performed a comprehensive experimental comparison between federated and centralized learning. We evaluated various classifiers on various datasets exploring influences of different sample distributions as well as different class distributions across the clients. ResultsThe results show similar performances under a wide variety of settings between the federated and central learning strategies. Federated learning is able to deal with various imbalances in the data distributions. It is sensitive to batch effects between different datasets when they coincide with location, similar as with central learning, but this setting might go unobserved more easily. ConclusionFederated learning seems robust to various challenges such as skewed data distributions, high data dimensionality, multiclass problems and complex models. Taken together, the insights from our comparison gives much promise for applying federated learning as an alternative to sharing data.

bioinformatics↗