Biology subjects

Zhang, S.

Publications and source records attributed to Zhang, S..

At least 73 records · Page 4Linked to original sources

bioRxiv2026

Mechano-Piezoelectric F-Peptide Hydrogels Enable In Situ Stem Cell and Immune Niche Modulation for Cartilage Regeneration

Osteoarthritis, characterized by cartilage degradation and synovial inflammation, has spurred interest in mechano-piezoelectric bio-hydrogel therapies that can both relieve symptoms and reverse progression. However, current approaches lack sufficient piezoelectric output and dual cartilage/inflammation targeting. To address this, we demonstrated a mechano-piezoelectric peptide hydrogel composed of nanofibers integrating piezoelectric cues with mesenchymal stromal cells (MSCs) recruitment and PIEZO2 mechanosignaling. Molecularly, the hydrogel's seed peptide incorporated four functions: COL2A1 targeting, MMP-13 responsiveness, MSCs homing, and self-assembly. Overexpressed MMP-13 in the osteoarthritis niche triggers gelation, promoting MSCs recruitment and drug retention. Fluorination modulates hierarchical nanofiber assembly, enhancing mechanical and piezoelectric properties, as confirmed by morphological, biophysical, and computational analysis. The trifluoromethyl-modified, 4-octyl itaconate (4-OI) loaded formulation reverses osteoarthritis via PI3K/AKT activation and Wnt/{beta}-catenin suppression, as shown by improved Osteoarthritis Research Society International (OARSI) scores, bone microarchitecture, and cartilage matrix. This synergy of mechano-piezoelectric cues and 4-OI offers a clinically promising strategy for osteoarthritis.

biophysics↗

bioRxiv2026

GWAS and multi-omics uncover the SLC6A3 locus for canine stranger-directed social anxiety and implicate NRIP3 in dopaminergic network regulation

The social anxiety disorder (SAD) is one of the most common mental health disorders, often developing in adolescence. It can severely impair educational achievement, career progression, and social functioning. Certain individuals within the Chinese Kunming dog (CKD) population display spontaneous, SAD-like behaviors towards strangers, offering a valuable model for genetic investigation. Using whole-genome sequencing data, we conducted a genome-wide association study on this SAD-like behavior in 100 CKDs, identifying several genes previously linked to human psychiatric disorders. Notably, SLC6A3 reached genome-wide significance, whereas CADPS2 exceeded only the suggestive threshold and is therefore considered a network-supported candidate locus. By integrating weighted gene co-expression network analysis with public RNA-seq data from the VTA of 38 mice (after quality control), we further showed that Slc6a3 and Cadps2 are part of a co-expression network module associated with the dopamine system. These findings suggest that this specific module may play a functional role in the stranger-directed anxiety-related behaviours in CKDs. Furthermore, we identified Nrip3 as a key candidate transcriptional coregulator in this specific module. Subsequent virus-mediated overexpression and behavioural experiments showed that Nrip3 regulates gene expression within this network and induces a spectrum of behavioural alterations in mice, including social avoidance, anhedonia, and increased locomotor activity. Collectively, this study highlights the involvement of dopamine system genes in canine social anxiety and suggests Nrip3 as an upstream regulator of Slc6a3 associated with anxiety. The identification of conserved anxiety-related signatures between humans and dogs further supports the dog as a valuable translational model for psychiatric research.

genetics↗

bioRxiv2026

MG2Act: A Mechanism-Inspired Sequential Attention Framework for Molecular Glue Degradation Prediction

Molecular glue degraders act by inducing productive proximity between an E3 ligase and a substrate protein. For most characterized degradative glues, a small molecule first engages the E3, conditions its substrate-recognition surface, and only then enables recruitment of a compatible neo-substrate. This directionality is rarely encoded explicitly in computational models, which typically fuse molecule, E3 and target representations simultaneously. We present MG2Act, a structure-independent framework that translates this two-step logic into sequential cross-attention, using CRBN-mediated degradation as the most data-rich representative system. Starting from a curated continuous-valued benchmark of 1,207 pairs across 47 targets, a refined subset of 1,159 pairs was selected to train MG2Act after excluding rare targets. On identical processed data, MG2Act consistently outperforms machine learning baselines, robustly generalizes under strict redundancy-filtering, and responds coherently to mechanism-based perturbations. Prospective screening and zero-shot target-conditioned prioritization identified nanomolar degraders of IKZF1, CK1 and CDK4, including the non-classical IMiD-core CDK4 degrader SWC-202.

bioinformatics↗

bioRxiv2026

A rationally designed neuraminidase immunogen elicits humoral responses to a conserved viral site

Efforts to develop a universal influenza vaccine have primarily focused on the surface-exposed viral hemagglutinin (HA), but neuraminidase (NA) is an additional target for cross-reactive, protective responses. Here, we used hyperglycosylation as an immunogen design approach to reshape anti-NA humoral immunity toward conserved antigenic regions. Iterative design produced a hyperglycosylated NA immunogen that retained enzymatic activity and reactivity to a conformation-specific antibody recognizing the conserved catalytic site. In mice, the hyperglycosylated immunogen elicited serum antibody responses of comparable magnitude to those elicited by the wild-type NA immunogen. However, serum antibodies elicited by the hyperglycosylated immunogen had increased breadth, recognizing N2 NAs from H2N2 and H3N2 viruses spanning nearly 65 years of antigenic drift, as well as a heterosubtypic N9 NA. Single B cell analyses identified a monoclonal antibody that competed with a component of the serum antibody response elicited by the hyperglycosylated immunogen, and structural characterization showed that it recognizes a previously undefined, conserved epitope at the NA tetramer interface. Passive transfer of this interface-directed antibody partially protected mice against lethal heterologous influenza challenge. Collectively, these data show that glycan shielding can reshape and enrich humoral responses toward a conserved antigenic region on NA. The immunogen and hyperglycosylation design strategy described here provide a template for developing next- generation NA-based influenza vaccines. ONE SENTENCE SUMMARYA hyperglycosylated influenza neuraminidase immunogen reshapes humoral immunity and elicits antibodies targeting a conserved tetramer-interface epitope.

immunology↗

bioRxiv2026

Plasma titration provides a physical ruler for cross-platform proteomics

Plasma proteomics is expanding across platforms and cohorts, and integrating these data for AI demands comparability at the protein level, not merely concordant associations1,2. Affinity and MS platforms use distinct probes (antibodies, aptamers, or peptides) and signal readouts, yielding contradictory cross-platform results3,4. Without a known quantitative truth, we cannot distinguish biology from measurement distortion, leaving no gold standard for integration. Here we introduce Plasmix--a plasma reference suite with predefined male:female ratios (M, 1:0; Y, 3:1; P, 1:1; X, 1:3; F, 0:1)--and show that preserving this quantitative titration gradient, not just technical repeatability, predicts cross-platform concordance and identifies protein measurements suitable for integration. Profiling Plasmix across five platforms (Olink, SomaScan, NULISA, AAgAtlas, and MS-DIA) and 12 protocols across 17 batches, we found discordance is dominated by signal generation, not sample identity, and platforms distort signals in a protein-specific manner. Crucially, proteins retaining the titration response showed stronger agreement in an independent cohort; anchoring to the Plasmix midpoint (P) via sample-to-reference ratios reduced distortions, extending harmonizable coverage by 10-20%. Plasmix thus provides a physical ruler to benchmark accuracy, identifying genuinely integrable measurements before pooling datasets or training AI models--a critical bottleneck for plasma proteomics.

molecular biology↗

bioRxiv2026

The chromatin reader protein MLLT1 is critical to maintain normal B lymphopoiesis

MLLT1 (also named ENL) is a chromatin reader protein whose encoding gene was originally identified as a chromosomal translocation partner with MLL(KMT2A) in acute leukemia. However, its role in normal hematopoiesis has not been investigated. This study uncovers a critical role of Mllt1 in normal B cell lymphopoiesis. We found Mllt1 to be essential for early B lymphocyte development using a conditional Mllt1 knockout mouse model that we developed. A significant decrease of bone marrow B-lineage progenitors, splenic transitional B cells and peripheral blood B cells were observed in Mllt1del mice compared to control Mllt1fl/fl mice. Similarly, Mllt1 deletion in in vitro cultured B-enriched progenitor cells from Mllt1fl/fl; Rosa26CreERT2/+ mice resulted in reduced B cells, demonstrating the cell-intrinsic role of Mllt1 in this process. Direct MLLT1 target genes including Il7r and critical B-lineage transcription factors, Ebf1 and Pax5, were decreased following Mllt1 deletion. Gene set enrichment, gene ontology, and functional analyses of Mllt1-deficient cells showed significant alterations related to B cell development, critical relevant signaling pathways, DNA replication, and mitochondrial function. In vitro complementation with MLLT1 rescued the B cell phenotype observed with endogenous Mllt1 deletion; however, specific MLLT1 YEATS domain mutants lacking chromatin reader and RNA-binding functions were unable to rescue the phenotype. Taken together, our research demonstrates a previously unappreciated role for MLLT1 as critical for maintenance of B cell lymphopoiesis.

cell biology↗

bioRxiv2026

Non-Invasive Embryo Quality Assessment via Matrix-Optimized Untargeted LC-MS Metabolomics of Spent Embryo Culture Media and Weighted Ensemble Machine Learning

BackgroundNon-invasive embryo quality assessment is a critical unmet need in assisted reproductive technology (ART). Preimplantation genetic testing for aneuploidy (PGT-A) is effective but requires invasive biopsy that may compromise embryo viability. Metabolomics of spent embryo culture media (SECM) offers a non-invasive alternative, yet analytical challenges--limited sample volume, high salt content, and abundant proteins--have hindered standardization and clinical translation. ResultsWe systematically optimized sample preparation for untargeted LC-MS metabolomics of SECM using human serum as a reference. Optimal conditions were highly matrix-dependent: SECM required 7x volume of 50% acetonitrile for extraction and 40% acetonitrile for reconstitution, whereas serum required 10x volume of 100% methanol and 100% water--reflecting that SECM contains more non-polar species than serum. Applying the optimized workflow to 120 clinical SECM samples (72 euploid, 48 aneuploid), we identified 102 differential metabolites between euploid and aneuploid embryos, with prominent enrichment of lipid pathways (fatty acid metabolism, {beta}-oxidation, sphingolipid metabolism) and involvement of amino acid (methionine, tryptophan) and TCA cycle metabolism. A weighted ensemble machine learning model discriminated aneuploid from euploid embryos with an AUC of 0.977, 100.0% specificity, and 89.6% sensitivity. Among 72 euploid embryos stratified by morphological grading (good, fair, poor), metabolic alterations progressed from mitochondrial energy deficiency (good vs. fair) to broader lipid dysregulation (fair vs. poor), with the ensemble model achieving AUCs of 0.944, 0.889, and 0.943, respectively. ConclusionsThis study establishes a rigorously optimized and validated SECM metabolomics workflow that overcomes key analytical barriers in this challenging matrix. Our findings demonstrate that metabolic signatures--particularly in lipid and energy metabolism--are strongly associated with both embryo ploidy and morphological quality, providing biological insights into the metabolic underpinnings of embryo developmental competence. The high predictive performance of the ensemble model supports the feasibility of non-invasive embryo assessment as a complementary tool to existing methods, with potential to reduce reliance on invasive biopsy in ART. External validation in prospective multi-center cohorts is warranted to further assess clinical utility and generalizability. Graphical Abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=122 SRC="FIGDIR/small/741666v1_ufig1.gif" ALT="Figure 1"> View larger version (59K): org.highwire.dtl.DTLVardef@10fbe40org.highwire.dtl.DTLVardef@936273org.highwire.dtl.DTLVardef@246e5dorg.highwire.dtl.DTLVardef@1c9d115_HPS_FORMAT_FIGEXP M_FIG C_FIG

genetics↗

bioRxiv2026

Synergistic cytotoxicity with Chk1/Chk2-inhibitor prexasertib in small cell lung cancer following lurbinectedin-induced G2/M-checkpoint activation

Small cell lung cancer (SCLC) is an aggressive thoracic malignancy with a 5-year survival rate under 7%. Lack of meaningful improvement of survival rates despite advances in treatment highlights the need for novel therapeutic approaches to improve patient outcomes. Currently, carboplatin + etoposide chemotherapy is the backbone of treatment for most patients. Lurbinectedin is a cytotoxic drug with unique activity against small cell lung cancers in patients with extensive disease and acquired resistance to carboplatin + etoposide. Our preliminary experiments in human SCLC cell lines treated with lurbinectedin demonstrated a dose-dependent increase in Chk1 and Chk2 protein phosphorylation. A consequence of the frequent TP53 inactivation in SCLC is tumor cell reliance on G2/M cell cycle checkpoints involving Chk1/Chk2 to maintain genomic integrity and allow cell survival following DNA damage. We hypothesised that inhibition of Chk1/Chk2-dependent responses with dual-inhibitor prexasertib (ACR-368), would potentiate tumor cell killing by lurbinectedin potentially in a synergistic manner. SCLC cells underwent cell death following single agent prexasertib exposure and this further increased with prexasertib + lurbinectedin combination. Highest Single Agent (HSA) synergy score calculations based on cell viability measurements suggested synergistic action between prexasertib and lurbinectedin at select dose combinations. Western blot analysis of intracellular proteins from SCLC cells treated with both drugs demonstrate dynamic, dose-dependent effects on Chk2, Chk1 and downstream effector Wee1, with lurbinectedin increasing intracellular levels of pChk1 and pChk2, while co-treatment with prexasertib deregulates this process across multiple human-derived cell lines. Synergistic killing was associated with elevated {psi}-H2AX levels indicative of DNA double strand breaks and PARP-cleavage due to apoptotic caspase activation. Despite some heterogeneity among treated SCLC cells, the increased phosphorylation of Chk1 was noted at several kinase-activating sites including Serine 296, 317, and 345 while Chk2 Tyrosine 68 phosphorylation was consistently upregulated by lurbinectedin. The results provide a preclinical mechanistic rationale for overcoming a pro-survival, drug resistance-promoting checkpoint pathway to enhance the unique efficacy of single-agent lurbinectedin in patients with SCLC.

cancer biology↗

bioRxiv2026

Structural bioinformatics of three Epstein-Barr Virus (EBV) Integral Membrane Proteins and their water-soluble QTY analogs

The Epstein-Barr virus (EBV) is a highly prevalent virus worldwide that is associated with several lymphoid and epithelial malignancies. However, extensive research on EBV integral membrane proteins BILF1, LMP1 and LMP2, has been scarce due to their hydrophobic transmembrane domains. Our study applies the QTY code (glutamine, threonine, tyrosine) to design water-soluble analogs of BILF1, LMP1 and LMP2 with reduced hydrophobicity, where we systematically replaced hydrophobic amino acid residues leucine (L), isoleucine (I), valine (V), and phenylalanine (F) with structurally similar polar residues glutamine (Q), threonine (T), and tyrosine (Y). We retrieved their native sequences from UniProt, identified transmembrane domains using Protter, then performed QTY design through the Protein Solubilizing Server (PSS). We then predicted native and QTY structures using in silico prediction tools AlphaFold3, ColabFold, and Boltz-2. Our analyses demonstrate that despite significant protein sequence replacements in their transmembrane domains (54.15%-61.59%) and increased intrinsic solubility, the QTY analogs exhibited minimal changes in isoelectric point (0.00-0.15 decrease) and molecular weight (0.7-1.2 kDa increase). Additionally, structural superpositions between QTY analogs and native structures using PyMOL yield low RMSD values (0.217[A] -1.202[A]). Our results demonstrate the QTY codes ability to design detergent-free analogs of BILF1, LMP1 and LMP2 with substantially reduced hydrophobicity and aggregation propensity whilst preserving native-like structures. Our results may facilitate protein characterization studies, therapeutic research on EBV, and other protocols that typically require protein solubilization.

bioinformatics↗

bioRxiv2026

GEM-GPT Enables Personalized Cell Type-Resolved Therapeutic Design for Systems Pharmacology

Generative AI is transforming drug discovery, yet most approaches follow one-drug-one-target paradigms ill-suited to the heterogeneity of chronic, systemic diseases. Systems pharmacology offers an alternative, but generative tools designed for it remain scarce. We introduce GEM-GPT, a transcriptomics-guided framework that generates personalized therapeutic candidate molecules intended to shift cell type-specific disease states toward healthy phenotypes. GEM-GPT uses a biology-inspired deep fusion architecture that couples a single-cell RNA-sequencing foundation model with a molecular GPT, modeling cell type-specific chemical-gene interactions throughout molecule generation rather than through fixed conditioning. Across bulk and single-cell chemical perturbations and CRISPR knock-out signatures, GEM-GPT outperforms state-of-the-art baselines, produces cell type-resolved molecules, and generalizes to unseen cellular contexts. In a case study on opioid use disorder (OUD), it generates novel candidates, recovers FDA-approved OUD-related drugs absent from training, and yields predicted binders to OUD-related targets. GEM-GPT bridges single-cell omics and molecular generation for personalized, cell-type-resolved, systems-aware therapeutic design.

pharmacology and toxicology↗

bioRxiv2026

Damaging RBM20 E-rich domain variants are not rescued by gene replacement

The promise of precision therapeutics in genetic cardiomyopathies relies on linking specific therapies to variant mechanisms. Missense variants in the cardiac splice regulator RBM20 cause a highly penetrant and arrhythmogenic dilated cardiomyopathy. Disease-causing variants in RBM20s arginine-serine rich (RS) domain act via formation of toxic gain of function cytoplasmic granules, but this is not true for a small number of clinically adjudicated pathogenic variants in its glutamate(E)-rich domain. To better define the effects of E-rich domain variants, we developed a scalable screen based on induced pluripotent stem cell (iPSC) cardiomyocyte differentiation that identified several additional damaging variants. Several of these reduced RBM20 protein abundance and stability. We therefore hypothesized that, unlike RS domain variants, these E-rich variants might be rescued by RBM20 overexpression. To test this hypothesis, we generated induced pluripotent stem cells (iPSCs) from a patient with a pathogenic E-rich domain variant (p.E913K), and confirmed reduced RBM20 protein expression in these RBM20+/p.E913K cells after differentiation to iPSC-derived cardiomyocytes (iPSC-CM, vs. engineered isogenic RBM20+/+). These iPSC-CMs also displayed aberrant transcriptional splicing, reduced contractility, increased calcium-induced calcium release, and nuclear localization of RBM20 protein, often with more than the two expected RBM20-centric splice factories. AAV-based overexpression of RBM20 reversed some, but not all of the mis-splicing events identified in RBM20+/p.E913K iPSC-CMs, and did not improve their abnormal contractility, calcium handling or supernumerary RBM20 nuclear granules. In summary, our data indicate that pathogenic E-rich domain variants reduce RBM20 protein abundance, but that their mechanism is unlikely to be explained by haploinsufficiency alone.

genetics↗

bioRxiv2026

Charged amino acid propensities and solubility: lysine is elevated at the termini of helices in E. coli proteins.

An emerging result in the relationship between amino acid sequence and protein solubility is a preference, on average, for lysine over arginine in more soluble proteins. The termini of helices are known to be prone to partial unfolding, often employing N- and C-cap amino acids to maintain stability. Hypothesising that lysine/arginine differences in relation to solubility may be evident at helical termini, their propensities and predicted charge interactions in helices were examined. There is enrichment of lysine over arginine at helical termini in AlphaFold models of Escherichia coli proteins, more so (on average) for the most soluble proteins. Similar effects are seen for the sum of charged amino acids at helical termini. Regions other than helical termini also show correlation of lysine composition, and overall charged amino acid composition, with solubility. These results suggest that protein design protocols could improve solubility through targeting lysine enrichment in regions such as helical termini, in addition to the more conventional consideration of helix capping interactions.

bioinformatics↗

bioRxiv2026

Telomeric G-Quadruplexes Formed by Six G-overhang Permutation Repeats

The single-stranded G-overhangs of vertebrate telomeres, composed of guanine (G)-rich hexanucleotide repeats, can adopt six register-dependent permutations: GGTTAG, GTTAGG, TTAGGG, TAGGGT, AGGGTA, and GGGTTA. Telomeric G-quadruplexes (G4s) formed by the first two repeat permutations, which are the most prevalent at human chromosome ends, have been largely overlooked. Here, G4 formation across all six permutations, each containing 4 to 9 repeats, are systematically investigated under varying cation conditions and molecular crowding environments. Our findings reveal that distinct repeat permutations yield markedly different G4 stabilities and folding topologies, which in turn modulate DNA polymerase activity. This expanded perspective on telomeric structural diversity provides new insights into telomere biology and suggests that multiple repeat patterns may cooperatively influence chromosome end protection and targeting.

biophysics↗

bioRxiv2026

Engineering the bacterial cellulose-forming surface as a programmable protein recruitment interface

Synthetic biology has advanced microorganisms to be programmed as production hosts, but its application to bacteria that inherently assemble extracellular materials remains limited. Komagataeibacter spp., natively synthesizes cellulose at the bacterial cell surface, creating a material-forming interface that has not been used as a programmable recruitment platform. Here we establish cell-surface display in Komagataeibacter intermedius and show that this interface can recruit defined proteins, making functionalization part of cellulose formation. By engineering LppOmpA, we displayed a fluorescent protein and genetically encoded capture modules (SpyTag and SilkTag) to selectively capture catcher-fused protein cargos onto K. intermedius cell surface. Recruitment of silk-derived structural protein before cellulose production generated silk-associated fibrous structures within the pellicles, with retained cargo signal after washing. The resulting biocomposite showed reorganized fibre-network morphology, increased surface hydrophobicity, mesoscale ordering, and improved wet-state compressive strength. Wild-type cells exposed to same conditions did not reproduce these changes, demonstrating that material properties arise from surface-directed recruitment rather than protein exposure alone. This work demonstrates the material-forming bacterial surface as a programmable engineering interface for organizing extracellular proteins, providing a general strategy for engineering living materials.

bioengineering↗

bioRxiv2026

A single chromosome 3p break initiates clear cell renal cell carcinoma evolution

Clear cell renal cell carcinoma (ccRCC) is initiated by chromosome 3p loss, yet chromosome losses impose a profound fitness burden on normal cells. How renal epithelial cells tolerate this deleterious aneuploidy during early tumorigenesis remains unclear. Analysis of 949 ccRCC genomes reveals two major classes of chromosome 3p alterations: simple deletions and complex rearrangements surrounding a terminal breakpoint - a pattern we term breakpoint-confined chromothripsis. We modeled both alterations in non-transformed human renal proximal tubule epithelial cells by introducing a single DNA double-strand break on chromosome 3p. Despite an initial fitness disadvantage, chromosome 3p loss drives adaptive genomic evolution that recapitulates recurrent ccRCC-associated aneuploidies, including 5q gain and 14q loss. These alterations alleviate the fitness constraints of 3p loss and promote metabolic reprogramming, clonal expansion, and malignant transformation, producing tumors with features of ccRCC. Thus, a single chromosome break initiates the evolutionary trajectory of ccRCC by creating a fitness bottleneck that selects for recurrent aneuploidies.

cancer biology↗

bioRxiv2026

Bone Mechanosensing Dictates Hematopoietic Stem Cell Fate and Immune Homeostasis

The vertebrate water-to-land transition was accompanied by a six-fold increase in gravitational force, followed by the migration of hematopoietic stem cells (HSCs) from kidney or liver to the bone marrow, and the acquisition of enhanced immune functions to cope with novel environmental pressures. Bone senses mechanical loading and provides a microenvironment for HSC development, yet whether bone mechanosensation affects immune cell development and immune homeostasis remains unclear. Here, we unveil bone as a mechanosensory organ that translates mechanical force into hematopoietic instructions. Mechanical loading of bone directs HSCs toward lymphoid lineages in mice and non-human primates, whereas unloading favors myeloid commitment. This process requires osteocyte mechanosensor Piezo1, which induces loading-responsive bone-derived factors such as IL1R2, SERPINC1 and INMT, thereby restraining inflammatory signaling and guiding HSC differentiation. Osteocyte Piezo1 deficiency recapitulates the hematopoietic and immune alterations observed during unloading. Functionally, this pathway enhances acute infection resistance and suppresses immunosenescence in mice and in aged long-tailed macaques. Notably, skeletal mechanoregulation of immune homeostasis is conserved across vertebrate species. Our research defines the mechano-bone-immune axis (mechano-osteoimmunology), offering a novel evolutionary perspective on the interconnected development of the skeletal and immune systems and presenting a promising non-pharmacological target to address immune dysfunction.

immunology↗

bioRxiv2026

Residual Multi-Modal Learning for Pan-Breast-Cancer Drug Response Prediction

Predicting drug sensitivity across diverse cancer cell lines remains a fundamental challenge in precision oncology, particularly for data-scarce cell lines where per-cell-line models overfit and lookup-table approaches cannot generalise to unseen biological contexts. We present DL4DR, a Two-Tower Residual Late Fusion deep learning model that addresses this challenge through content-based, identity-free genomic conditioning. The Cell Line Tower encodes each cell line as a 3 x 139 x 139 genomic image - encoding gene expression, mutation severity, and copy-number variation as RGB channels - using a convolutional encoder that maps directly from biological content, never from a cell line ID. The Compound Tower combines three complementary molecular representations: D-MPNN graph message passing, ORNN octave convolutional image features, and an ECFP hard-memorization head that preserves activity-cliff resolution. Predictions are composed as a residual sum: f = fhard +{lambda} (zC) {middle dot} fresidual, where the learned gate{lambda} modulates how much interaction signal supplements the memorization baseline. Evaluated across 51 breast cancer cell lines (136,342 records), Residual Fusion outperforms the ECFP-Only baseline in 48/51 cell lines (94.1%), with {Delta}R2 > 0.02 in 26/51 (51.0%). On the leave-cell-line-out split - the decisive test of genomic generalisation - the mean {Delta}R2 = +0.016 across all 51 lines demonstrates that the genomic encoder learns transferable biological signal beyond cell line identity. External validation on 601 cell lines across 27 cancer tissue types (CellTiter-Glo dataset; 0 cell line overlap with training) achieves median R2 = 0.627, within the range of the internal random-split performance (R2 = 0.61-0.69), confirming pan-cancer generalisation. GradCAM interpretability on the Cell Line Tower recovers TP53 among the top-five cross-cell-line genomic activators (5/51 cell lines) alongside several uncharacterised candidate genes (e.g. FSIP2, 6/51) - without any prior pathway annotation - providing partial biological validation of the learned representation, while also indicating that a substantial share of the encoders top-ranked signal corresponds to genes with no current annotation as breast cancer drivers. Code and data are available at https://github.com/bayjuan5/DL4DR.

bioinformatics↗

bioRxiv2026

Consistency Analyses of Open-source Software for Motor Unit Decomposition Using High-density Electromyography Signal

Motor unit (MU) decomposition using high-density surface electromyography (HD-sEMG) has been widely used to characterize MU behavior in neurophysiology and to build neural-machine interfaces for wearable robots. Recently, many open-source software tools for MU decomposition have been made available on GitHub, which could reduce the effort of researchers in the field. However, the consistency among these open-source tools has never been studied, making researchers hesitate to use them. In this study, we collected 7 open-source software tools on GitHub and applied them to decompose MUs from an open-source HD-sEMG dataset (including 11 isometric contraction trials) to investigate the consistency among these tools. To create a comprehensive MU pool for reference, we combined all unique MUs identified by seven tools, visually inspected and removed bad MUs, and manually edited all remaining MU spike trains. Across 7 tools for 11 trials, the number of identified MUs ranges from 167 to 736. The number of valid MUs after expert inspection ranges from 29 to 210, which is 10% to 72% of the reference pool. The rate of agreement between the raw MUSTs and the manually edited MUSTs ranges from 0.86 to 0.94, and the averaged number of edits per MU to correct misalignments ranges from 14 to 39. The results show inconsistency in the implementation and procedures of each tool, which results in an inconsistent number of identified MUs and valid MUs (29 vs 210). In general, a substantial amount of effort is required to process the raw MUSTs from each tool to conduct further research analysis. This study provided a guideline for using open-source software tools for MU decomposition and indicated that it would be beneficial to develop tools to automatically edit the MUSTs.

bioengineering↗