Immunogenetics

Blood group antigens, defined by epitopes on the erythrocyte surface, are central to transfusion safety and maternal-fetal compatibility. While the genetic basis of many clinically relevant blood group antigens is well established, which structural and biophysical parameters determine whether a single-nucleotide variant gives rise to an antigenic phenotype remains unclear. Here, we integrate structural, biophysical, and evolutionary analyses to systematically evaluate features associated with single amino acid substitutions across 24 human protein-based blood group systems. We analyse 319 variants with curated phenotypic annotations alongside 481 control variants, identifying key determinants of null and antigenic phenotypes. Null variants are characterized by high evolutionary conservation, burial within the protein core, loss of hydrophobicity, increased polarity, and a propensity for arginine substitutions. Antigenic variants are also enriched in arginine; however, in contrast to null variants, they tend to occur at less conserved, more solvent-accessible, and structurally flexible sites. Supervised machine learning models trained on structural and biophysical descriptors were applied to distinguish (i) null and (ii) antigenic variants from controls, achieving balanced accuracies of 0.82 and 0.63, respectively. Feature importance analysis identified predicted pathogenicity, solvent accessibility, and evolutionary conservation as the most predictive determinants of null variants, whereas hydrophobicity, conservation, and flexibility dominated antigen prediction. This work establishes a framework linking molecular variation to blood group phenotypes and provides a foundation for predicting the impact of novel missense mutations in transfusion medicine and beyond.

Background Complex traits are influenced by numerous variants, most of which have regulatory effects on gene expression that can be mediated by DNA methylation. Molecular QTL mapping is an approach that aims to dissect these effects. However, obtaining molecular phenotypes on a large scale is challenging, particularly in livestock species. In cattle, an epigenotyping array called EpiChip has recently been developed in the European RUMIGEN project. The EpiChip, which contains 43,317 CpG sites distributed all over the bovine genome, enables large-scale measurement of DNA methylation. This study aims to characterize the genetic determinism of blood DNA methylation in cows by estimating heritability and mapping cis- and trans-methylation QTLs (meQTLs). Results Whole blood samples from 4,457 genotyped Holstein cows were epigenotyped. Across all CpG sites, the heritability estimates averaged 24.6%. The local meQTL mapping at sequence-level for variable CpG sites (SD > 2.5%; n = 28,806) detected cis-meQTLs for 80.1% of the CpG sites, with sentinel SNPs located close to their associated CpGs. A two-step analysis was also conducted to identify long-range associations, with a particular focus on trans-meQTL hotspots. First, we identified CpG-SNP trans-associations using medium-density genotypes (50k SNPs) that revealed 31,846 SNPs with significant effects on 1 to 530 trans-CpG sites. Then, regions associated with at least 34 independent trans-CpGs were retained defining 31 hotpots. For each hotspot, a local sequence-level GWAS was conducted using the first principal component derived from the associated trans-CpGs. Out of the 31 detected hotspots, three were located close to transcription factor genes (RUNX1, NFIC and FOXA3) for which the associated trans-CpGs were enriched for the corresponding binding motif. Two other hotspots were located within KDM5A and KDM5B, and their corresponding trans-CpGs were strongly overrepresented in H3K4me3 narrow peaks in blood as well as in other tissues. Conclusions By identifying functional candidate genes associated with blood DNA methylation in cattle, these findings provide new insights into the regulatory architecture of DNA methylation in mammals, highlighting the value of large-scale molecular data from livestock populations.

Pancreatic cancer remains a highly lethal malignancy with limited therapeutic options. Chimeric antigen receptor (CAR) therapy has revolutionized the treatment of hematological cancers but still faces major limitations in solid tumors, particularly due to the scarcity of tumor-specific targets. Cutaneous lymphocyte antigen (CLA) recently emerged as a promising PDAC target due to its high tumor expression and limited presence in healthy tissues. However, previously reported CLA-directed CAR constructs lacked antitumor functionality. Here, we investigated multiple strategies to generate functional CLA-targeting CAR approaches. We first hypothesized that impaired activity resulted from fratricide caused by CLA expression on activated T cells. CLA knockout was successfully achieved through deletion of fucosyltransferase-7, but not by knockout of the major CLA carrier backbones CD162, CD44 or CD43, suggesting additional CLA carriers or compensatory regulation. As CLA knockout alone did not restore CAR-mediated killing, we explored whether insufficient binding affinity limited CAR activity. Affinity maturation was performed in silico and in vitro using yeast surface display, identifying 39 candidate mutations, although none restored cytotoxicity. We finally switched to an AdCAR strategy using anti-biotin CAR T cells combined with biotinylated anti-CLA scFv-Fc adapters. This approach enabled efficient, concentration-dependent cytotoxicity with both CLA-targeting binders. Additionally, we identified a dynamic, cell density-dependent regulation of CLA expression. Finally, glycan profiling of CLA binders further revealed broader-than-expected glycan interactions, suggesting a potentially wider definition of the CLA family. Overall, our findings establish CLA as a functional PDAC immunotherapy target while revealing unexpected complexity in its regulation and molecular presentation.

Background: Abdominal aortic aneurysms (AAA) are characterized by dilation of the aorta that can lead to aortic rupture and death. The transcriptional co-activators Yes-Associated Protein (YAP) and WW-domain-containing transcriptional co-activator with PDZ-binding motif (TAZ) are mechanosensitive effectors of the highly conserved Hippo signaling pathway. It is hypothesized that cell-specific YAP/TAZ signaling in endothelial cells (EC) plays a pivotal role in mediating AAA formation and rupture. Methods: Single-cell RNA-sequencing in human AAAs was performed and differentially expressed genes (DEGs) were identified in the endothelial cell cluster. YAP/TAZ mRNA and protein expression were also assessed in human AAA and control aortic tissue. Two established murine AAA models were used with male C57BL/6 and EC-CreERT2-YAPfl/fl/TAZfl/fl mice with/without Verteporfin (VPF, YAP/TAZ inhibitor) and XMU-MP-1 (YAP/TAZ activator) treatments. On postoperative days 14 and 28, aortic diameter, histology, cytokine, and MMP2 expressions were evaluated. Results: A significant alteration in EC-specific differentially expressed YAP/TAZ-related genes was observed in which 242 genes were upregulated and 71 genes were downregulated in AAA compared to controls. Human AAA tissue showed a significant increase in YAP and TAZ protein expressions compared to controls. Elastase-treated EC-YAP/TAZ-/- mice showed a significant decrease in AAA diameter compared to littermate controls. Histological quantification revealed preservation of -smooth muscle actin, reduced elastin fiber breaks, and decreased macrophage infiltration in EC-YAP/TAZ-/- mice compared to littermate controls. Importantly, pharmacological inhibition of YAP/TAZ using VPF significantly attenuated AAAs in two experimental murine models. In vitro data demonstrates that VPF inhibits endothelial cell YAP expression, downregulating pathways associated with pathogenic angiogenesis and vascular inflammation. Conclusions: These data suggest that EC-specific YAP/TAZ signaling mediates AAA formation. Pharmacological inhibition of the Hippo pathway can significantly mitigate aortic inflammation and vascular remodeling to decrease the progression of AAAs and prevent aortic rupture.

The regnase family of endoribonucleases regulates immune gene expression through degradation of cellular mRNAs. Regnase-1 is known to also target viral RNA, but whether antiviral activity is a conserved property of all human regnases remains unknown. Here, we systematically compare the antiviral properties of all four human regnases. Regnases-1-4 expression inhibited HIV-1, HIV-2, MLV, RSV and hCoVs OC43 and SARS-CoV-2, but not the DNA virus HSV-1. Endogenous knockout and knockdown experiments demonstrated that physiological regnase expression restricts HIV-1 replication in a cell-type-dependent manner. Regnase-1 and regnase-4 were induced by interferons in macrophages, and all four regnases displayed signatures of positive selection during mammalian evolution, consistent with their potential roles as antiviral restriction factors. Mechanistically, antiviral activity of regnases required intact catalytic core and CCCH zinc finger domains, while nuclear shuttling and dimerisation site conservation were not shared features of all family members. Domain-swap and reporter analyses further showed that differences in antiviral potency between regnases primarily reflect differential RNA target recognition rather than catalytic activity. Regnase-1 exhibited broad RNA targeting, whereas regnases-2-4 displayed more selective targeting profiles. Collectively, our findings establish the human regnase family as evolutionarily diversified antiviral RNA restriction factors with distinct substrate specificities.

Pathogenic variants in COL3A1 cause Vascular Ehlers-Danlos syndrome (vEDS), a rare connective tissue disorder characterised by vascular fragility, increasing the risk of arterial ruptures/dissection. Advances in genomic sequencing have led to an increasing number of COL3A1 variants where the clinical significance is unclear, with these being termed variants of uncertain significance (VUS). VUS creates challenges for diagnosis and clinical management. Thus major efforts have been made to reclassify these to either pathogenic or benign variants in disease causality. Functional data from model systems can provide significant evidence to clinicians on the pathogenicity of a variant. To address the increasing numbers of VUS in COL3A1, we developed a fast pipeline using F0 crispant zebrafish to provide functional evidence for variant classification despite there being no direct orthologue of COL3A1 in zebrafish. Loss of col5a1 resulted in cardiac defects, dysmorphic blood vessel structures and delayed angiogenic sprouting. Trunk haemorrhage prevalence under physical stress increased in col5a1 knockout zebrafish, recapitulating vEDS patients. Remarkably, co-injection of F0 col5a1 knockout crispants with human wildtype COL3A1 mRNA partially rescued cardiac and vascular phenotypes, indicating a level of functional conservation between zebrafish type V and human type III collagen. These findings establish a tractable in vivo platform for functional assessment of COL3A1 VUS. Phenotypic rescue with wildtype COL3A1 provides a benchmark against which the pathogenicity of variants can be evaluated, generating functional evidence for VUS reclassification. Our model provides both a valuable tool for investigating vEDS disease mechanisms and a clinically relevant platform to improve diagnoses for patients with suspected vEDS.

Transcriptome assembly enables the recovery of expressed genes and isoforms, but the optimal strategy for reconstructing transcriptomes from long-read sequencing remains unresolved. In particular, establishing best practices for generating accurate gene models and selecting representative isoforms is essential for comparative genomics, as for orthology inference typically only the longest isoform per gene model is included. Here, we systematically compare clustering and de novo assembly methods using PacBio Iso-Seq data from 17 animal lineages spanning seven phyla, most of them non-model species, with the goal of investigating which methodology is more adequate to select one isoform per gene model, in the absence of specific pipelines to do so. We evaluate four approaches: isoseq cluster, CD-HIT, RNA-Bloom2 and isONform. We benchmark them with short-reads using Trinity, assessing assembly quality with BUSCO completeness, short-read mapping rates, coding sequence recovery, and longest isoform prediction. Our results show that CD-HIT clustering at high similarity thresholds ([≥]99%) yields the most complete and coding-rich long-read transcriptomes, rivaling Trinity while avoiding its high redundancy. Consensus-based methods such as isoseq cluster and isONform recover fewer single-copy orthologs (mirrored in a lower BUSCO score) and achieve lower mapping rates, while RNA-Bloom2 provide intermediate performance with reduced duplication. Together, these findings establish, to date, CD-HIT as a robust and practical strategy for transcriptome reconstruction from long-read data when genomic references are unavailable. By benchmarking de novo methods across a taxonomically broad dataset, this work defines the realistic capabilities of long-read transcriptome reconstruction in the absence of a reference genome and provides practical guidance for deriving high-quality gene models and selecting representative isoforms for orthology inference in non-model species.

Immunoglobulin D (IgD) is among the most conserved antibody isotypes, found in virtually all jawed vertebrates. Unlike other isotypes, IgD contains an unusually long hinge region of up to 160 amino acids that connects the constant and variable regions. Its expression pattern is also conserved; IgD is co-expressed with IgM on transitional and mature naive B cells. However, the function of IgD has remained enigmatic since its discovery in 1965. Here we present and test a biophysical model positing that IgD increases the entropic cost of bivalent antigen binding. Single-molecule measurements revealed that the antigen-binding arms of IgD are substantially more dynamic than those of IgM, suggesting that cell surface IgD would be energetically penalized in bivalent antigen binding. Consistent with the model and biophysical data, we find that the long hinge compromises antigen capture by IgD B cell receptors (BCRs) compared to IgM BCRs. To determine how the difference in antigen binding impacts immunity, we produced mice that express only IgM and IgD, exclusively IgM or IgD, or IgD with a truncated hinge region. The data indicate that the increased entropic cost of antigen binding imposed by the IgD hinge attenuates negative selection by self-antigen while increasing the affinity-based threshold for positive selection into the germinal center (GC). Together the results indicate that IgD functions physiologically to desensitize B cells to antigen, thereby expanding the B cell repertoire while optimizing affinity-based selection into the GC.

Spectral bleed-through remains a persistent practical problem in multichannel fluorescence microscopy. Signal from one fluorophore can be recorded in the detection channel of another, thereby biasing intensity measurements, inflating apparent colocalization, and complicating the interpretation of dynamic microscopy data. Although many correction strategies exist, routine workflows often remain fragmented across ad hoc scripts, manually tuned graphical procedures, or method-specific blind-unmixing implementations with limited provenance. Here we present spectral-unmixing, an open-source Python package for reproducible linear spectral unmixing in multidimensional microscopy stacks. The package unifies directed two-channel correction with multiple alpha-estimation strategies, optional bidirectional two-channel correction through explicit inversion of a 2 x 2 mixing model, and PICASSO-family blind unmixing for multichannel data. Microscopy inputs are normalized at the API boundary to canonical TZCY X stacks, allowing the same unmixing code to be applied across file formats without manual axis handling. Machine-readable sidecar reports preserve the effective processing configuration and estimated coefficients for every output, so that workflows can be audited and reproduced. Synthetic and real-data-derived benchmarks show that the implemented workflows accurately estimate and correct bleed-through when their model assumptions are satisfied. In fixed-alpha two-channel simulations, the mean-ratio and linear-fit estimators recovered {approx} 0.283 for a ground-truth value of 0.28 and reduced target-channel normalized root mean squared error from approximately 0.029 to 0.003. In time-varying simulations, per-time-point estimation tracked coefficient drift substantially better than reference-time-point estimation. Bidirectional inversion recovered reciprocally mixed channels accurately when coefficients were known or well estimated. PICASSO-family benchmarks further showed a practical trade-off between reducing residual inter-channel dependence and preserving fluorophore identity, with MATLAB-style workflows behaving more conservatively and source-sink formulations providing stronger dependence suppression when meaningful directional priors are available. Together, these elements make spectral-unmixing a practical, transparent, and extensible platform for reproducible spectral unmixing of fluorescence microscopy data in neuroscience and other quantitative bioimage-analysis settings.

Temporal niche partitioning is a strategy for reducing interspecies competition and strengthening reproductive isolation. It relies on animals confining their daily activity to distinct diurnal, crepuscular, or nocturnal windows. However, a hardwired temporal niche is only advantageous under stable, predictable ecological regimes; surviving dynamic environments demands behavioral flexibility. Yet, it remains unclear how animals override rigid biological constraints to rapidly exploit transiently available fitness-critical time windows. To address this, we leveraged the twilight-active, species-rich Drosophila genus and monitored their daily activity under naturalistic conditions. Here, we show that intense sociosexual interactions rapidly drive a species-specific reformatting of their canonical crepuscular niche. The dominant sensory modality used for sexual communication predicts niche shift direction: reliance on chemosensation for courtship redirects behavioral activity into the night, while visual reliance shifts it into the day. This temporal plasticity bypasses the circadian clock, instead operating via a conserved dopaminergic pathway. Dopamine operates a dual-output brain circuit that simultaneously inhibits sleep and sustains sexual motivation. Our results reveal how mating imperatives decouple behavioral timing from circadian command, enabling conditional colonization of otherwise restricted temporal windows. Ultimately, by driving the divergence of previously overlapping niches, sociosexually induced temporal plasticity provides a powerful mechanism for sympatric coexistence in crowded environments.

The incidence of early-onset colorectal cancer (CRC) has risen sharply in recent decades1, yet the biological basis underlying the distinct behavior of tumors arising in young versus aged tissues remains poorly understood. Here we show that aging reprograms the epigenetic landscape of the colon, restricting colon tumor growth through stable silencing of developmental and fetal gene programs. We find that colon tumors arising in aged mice are intrinsically less proliferative than those arising in young animals. Multi-omic profiling of normal colon and colon tumors reveals that aging drives DNA hypermethylation, loss of Polycomb-associated chromatin states, and reduced chromatin accessibility at a defined set of developmental genes that are bivalent (marked by both H3K27me3 and H3K4 methylation), transcriptionally active in colon tumors from young animals and repressed in both tumors and normal tissue from old animals. Among the genes most strongly repressed in old animals is Tacstd2 (Trop2), a regulator of fetal intestinal programs and epithelial stemness. Pharmacologic inhibition of DNA methylation reactivates the aging-silenced gene network in organoids from old animals, whereas genetic disruption of Tacstd2 suppresses growth and developmental transcriptional programs in young tumor organoids. TACSTD2, fetal gene signatures, and the aging-associated bivalent gene program are likewise repressed in late-onset vs. early-onset human colorectal cancers. Collectively, these findings identify age-associated epigenetic silencing of developmental gene programs as a causal mechanism that constrains colorectal tumor growth and provide a mechanistic framework for understanding the distinct biology of early-onset colorectal cancer.

Extrahepatic delivery of lipid nanoparticles (LNPs) to non-phagocytic cells is a major challenge, with the leading strategy involving surface functionalization with target-specific monoclonal antibody (mAb) ligands. We investigate the stability of mAb-conjugated LNPs using two anchoring systems: the commonly used DSPE-PEG2kDa-maleimide and a block copolymer, PCL5kDa-b-PEG2kDa -maleimide, with the hypothesis that conjugation to a 150,000 Da antibody could overwhelm the relatively small ~600 Da aliphatic anchor on the PEG-lipid in vivo. Shedding of the mAB would compromise targeting. Conjugation integrity following IV injection was assessed by tagging LNPs and mAbs with metal ion tracers that could be quantified by ICP-MS. Results show that DSPE-PEG-mAb rapidly (within 1h) dissociates from LNPs in blood, leading to accelerated LNP clearance. In contrast, mAbs conjugated using PCL-b-PEG remained stably associated with the LNP over the 24h circulation and clearance of the construct. Results are connected to a thermodynamic model that reproduces experimental findings for PEG-anchor(-mAb) shedding in vitro and in vivo. This study identifies anchoring strength as a critical, unconsidered parameter for in vivo performance when conjugating mAbs to LNPs for extrahepatic delivery.

Blockade by sedimentary particles, such as mineral crystals, is a continuous risk the kidney tubule faces. To prevent that, kidney resident macrophages form transepithelial protrusions and remove intratubular sedimentary particles, a behavior particularly prevailing in the medulla over the cortex. However, the molecular mechanisms underlying this characteristic behavior of medulla macrophages are incompletely understood. In this study, we identified that the medulla had higher mechanical stiffness than the cortex in steady state, which was further elevated when kidney stone formed. Increased tissue rigidity was sensed by medulla macrophages via mechanoreceptor Piezo1, which promoted macrophage protrusion formation and their ability to clean the tubules. Loss of Piezo1 expression in kidney macrophages predisposed mice to intratubular accumulation of mineral crystal in steady state and accelerated kidney stone formation during oxalate intake challenge. Signaling via Piezo1 mobilized molecules involved in cell adhesion and protrusion assembly, including Talin2 and focal adhesion kinase (FAK). Finally, we developed a first-of-its-kind cell-based therapy for the treatment of experimental nephrolithiasis by exploiting macrophage Piezo1 activity, and this strategy shows great promise for future translational research.

Single-nucleus transcriptomic atlases offer an unprecedented opportunity to connect cellular molecular states with Alzheimer's disease (AD) neuropathology, but whether these profiles encode reproducible, predictive information about pathological burden remains unclear. We present the SEA-AD DREAM Challenge, an open, international, model-to-data competition built on the Seattle Alzheimer's Disease Brain Cell Atlas to predict Alzheimer's disease neuropathological severity from single-nucleus RNA-sequencing data. Participants developed containerized models to predict categorical neuropathological staging, including overall Alzheimer's disease neuropathologic change, Braak stage, Thal phase, and CERAD score, as well as quantitative amyloid-{beta} and phospho-tau burden measured by 6E10 and AT8 immunohistochemistry. Across 17 eligible teams from 15 countries, the crowdsourcing framework enabled systematic comparison of diverse computational approaches and surfaced a broad landscape of modeling strategies and candidate predictive features. Top-performing methods achieved near-perfect prediction of categorical staging, with the best submission reaching a quadratic weighted kappa of 1.0 for the Overall AD Neuropathological Change score (ADNC), and competitive prediction of quantitative pathological burden in held-out data, with a best concordance correlation coefficient of 0.48. Post hoc perturbation analyses revealed that top categorical-stage predictions relied heavily on donor-level metadata-driven signals rather than transcriptomic features, whereas quantitative pathology prediction was more robust and supported by transcriptomic and cell-type-associated features with potential biological relevance to AD progression. The challenge also introduced the first AI Agent Track in a DREAM Challenge, providing an early benchmark for autonomous and human-guided agentic model development in single-cell neuroscience. This work demonstrates that single-nucleus transcriptomes encode substantial information about Alzheimer's disease pathology, establishes a reproducible benchmark for molecular neuropathology prediction, and highlights critical principles for designing privacy-preserving, leakage-aware community challenges using deeply phenotyped human brain data.

Social media users often feel that time vanishes while scrolling, but real feeds confound novelty, rewards, social signals, and self-paced control, leaving the driver of this distortion unclear. We tested whether self-paced visual exploration is sufficient to compress subjective time by comparing active scrolling with passive, yoked viewing and a static baseline. Twenty-three adults viewed sequences of natural images under three within-subject conditions: Scrolling (self-paced mouse clicks), Watching (a passive, yoked replay of their own scrolling sequence), and a Baseline (a static image). Participants estimated the elapsed duration of each block. Subjective duration was most compressed under Scrolling (48% of elapsed time), followed by Watching (51%) and Baseline (65%). Two sources separated these effects. Adding back the empty inter-image fixations brought the image-rich conditions to within seconds of the Baseline, showing that observers barely counted the blank gaps; the Scrolling--Watching difference, by contrast, was independent of these shared gaps, isolating self-paced control as a second source of compression. Electrophysiology linked that control to anticipatory neural states and the timing of early visual responses, with no amplified encoding of individual images. The results favor an attention-weighted account of timing, on which subjective duration tracks how much attention reaches the clock, a resource that a self-paced stream and its uncounted gaps both draw away.

We showed before that the endogenous peptide Nociceptin blocks the premature differentiation of oligodendrocytes (OLGs), preventing untimely precocious myelination in the developing brain. Consistent with this early function, Nociceptin brain expression is developmentally regulated, sharply decreasing with the initiation and progression of myelination. However, we now found that at difference with controls and relapsing-remitting multiple sclerosis (RRMS), Nociceptin levels are highly elevated in cerebrospinal fluid from patients with the most severe progressive MS (PMS) forms. This questioned whether Nociceptin early developmental effects could be latter recapitulated, interfering with remyelination in PMS. This possibility was tested by inducing experimental autoimmune encephalomyelitis in older mice, at an age equivalent to that with increased risk of RRMS transition into PMS. Older animals develop persistently highly debilitating clinical symptoms, and display both brain and spinal cord demyelination. Importantly, these mice exhibit elevated brain Nociceptin levels, and their treatment with an antagonist of the Nociceptin receptor (NOR) elicits a regression of clinical scoring that is accompanied by higher ratios of OLGs/OLG progenitor cells, increased myelination, and reduction of reactive astrocytes. These findings suggest that Nociceptin may be a crucial player in the age-related progression of MS; interfering with OLG maturation and remyelination, and perhaps further exacerbating neurological dysfunction by targeting astrocyte populations. The upregulation of Nociceptin secretion by human astrocytes in response to proinflammatory cytokines, also points to this peptide as a mediator of microglia-astrocyte interactions supporting MS progression with aging. NOR may offer a novel pharmacological target for ameliorating the devastating effects of MS progression.

Medical foundation models compress biomedical data into embeddings that support diverse downstream clinical tasks. However, successful model deployment is hampered by performance degradation on external data. It is recognized that embeddings capture acquisition signatures, such as hardware and technical differences, in addition to biology. Effective harmonization must remove the acquisition signature while preserving biological signals, a trade-off that current methods fail to balance adequately. Input-level normalization fails to eliminate acquisition signatures from embeddings, whereas embedding-level methods adjust features in an untargeted manner. We present FEATMAP, a harmonization approach that models acquisition signatures as geometric distortions between manifolds of similarly arranged embeddings. Using paired data that isolate the effect of acquisition signatures, FEATMAP fits a single global affine transformation per foundation model to correct acquisition signatures directly in the embedding space. This targeted, reusable correction aims to preserve biological and demographic variation while harmonizing across acquisition signatures. Across scanner and foundation-model harmonization in digital pathology and field-strength harmonization in brain MRI, FEATMAP improves cross-condition embedding similarity, reduces performance gaps without retraining, and suggests potential for the alignment of disparate embedding spaces.

Heterochromatin protein 1 (HP1 is an intrinsic component of heterochromatin domains where it is involved in a diverse set of functions including heterochromatin spreading and organization, chromatin compaction and transcriptional silencing. It has been suggested that HP1 functions through a phase separation mechanism, a process that has been observed in vitro in the presence of N-terminal phosphorylation, nucleic acids and nucleosome arrays. HP1 can also interact with numerous binding partners that contain a specific motif called an HP1 access code (HAC). HACs recognize and bind to an interface formed by the chromoshadow (CSD) domains in the HP1 homodimer, the functional form of the protein. It has been shown that some HP1 binding partners can enhance its phase separation ability while others disrupt the process. Here, we focus on the interactions between HP1 and three binding partners, namely the p150 subunit of the chromatin assembly factor 1 (CAF-1), the N-terminal domain of the lamin B receptor (LBR), and the mitotic protein Shugoshin 1 (Sgo1). Using phase separation assays, we show that CAF-1 prevents HP1 phase separation while LBR and Sgo1 enhance it. Binding assays, mutational studies, NMR spectroscopy and computational analysis allow us to dissect the contributions of the HAC motifs, the charge patterns of the binding partner sequences and the role of N-terminal phosphorylation on HP1 in condensate formation. Our results demonstrate that each binding partner uniquely balances these contributions to modulate the properties of HP1, while electrostatic interactions dominate the regulation of phosphorylated HP1. These results suggest that HP1 binding partners play an important role in the modulation of its properties and the regulation of its functions in distinct biological contexts.

Brain areas differ in their inherent susceptibility to focal seizures, but the principles governing this risk remain unclear. While prior work has focused on anatomical and physiological factors, here we observed a fundamental contribution from the computations performed by the underlying neural network. Handcrafted and trained recurrent neural networks supporting continuous representations respond to seizure perturbations with higher activity and earlier performance decline relative to matched networks stabilizing discrete, well-separated states. Consistent with this prediction, in vivo recordings revealed that medial entorhinal cortex, whose grid cells exhibit continuous attractor dynamics, drives acute epileptiform discharges with stronger involvement and smoother state trajectories compared to CA3, a hippocampal subfield associated with discrete memory storage. Moreover, selective synaptic silencing demonstrated that this difference in seizure responses depends on intact entorhinal connectivity. Thus, the computations that enable neural networks to process information also influence their vulnerability to pathological transitions.

Transformer-based deep learning models have shown great potential for decoding visual EEG signals. However, their internal attention mechanisms are often evaluated primarily on optimization objectives, leaving their alignment with biological brain connectivity an open question. This study empirically evaluates how variations in EEG preprocessing strategies affect these attention representations using the Adaptive Thinking Mapper (ATM) model as a framework. We compared a baseline pipeline (MVNN only) against a comprehensive cleaning pipeline integrating ICA and notch filtering. The models were evaluated through cross-generalization, noise robustness, and spectral-temporal ablation analyses. Furthermore, we investigated the structural correspondence between the model's data-driven attention weights and neurophysiological reference networks (GPDC, PDC, and DTF) using Node Strength Correlation and Representational Similarity Analysis (RSA). The results show that the comprehensive preprocessing successfully suppresses non-neural artifacts, such as frontal noise and electrical interference, while maintaining comparable decoding accuracy and baseline robustness. Alignment analyses revealed that the broad spatial organization of the learned attention patterns remains highly stable across pipelines, capturing key directed connectivity dynamics with subtle, metric-dependent variations in global representational geometry. This work provides an empirical exploration into bridging data-driven attention weights with neurophysiological consistency, offering insights toward more transparent brain-computer interfaces.