Biology subjects

Kassis, T.

Publications and source records attributed to Kassis, T..

12 recordsLinked to original sources

bioRxiv2026

MetaMuse: A Multi-Agent AI System for Biomedical Metadata Curation and Harmonization

Inconsistent and unstructured metadata in public biomedical repositories, such as the Gene Expression Omnibus (GEO), severely limits data discoverability and research reproducibility. To address this, we introduce MO_SCPLOWETAC_SCPLOWMO_SCPLOWUSEC_SCPLOW, a modular, multi-agent artificial intelligence framework designed to autonomously extract, validate, and standardize unstructured biomedical metadata. Operating through a three-stage architecture utilizing large language model agents, specialized CO_SCPLOWURATORC_SCPLOWAO_SCPLOWGENTSC_SCPLOW contextually extract candidate values for specific target metadata fields. A centralized AO_SCPLOWRBITRATORC_SCPLOWAO_SCPLOWGENTC_SCPLOW enforces cross-field logical consistency to prevent contradictory annotations. Finally, a NO_SCPLOWORMALIZERC_SCPLOWAO_SCPLOWGENTC_SCPLOW leveraging a domain-specific semantic search model (SapBERT) maps these free-text candidates to formal ontological terms. We evaluated MO_SCPLOWETAC_SCPLOWMO_SCPLOWUSEC_SCPLOW on a gold-standard dataset of manually curated GEO samples, achieving over 95% curation accuracy across key target metadata fields, and demonstrated robust scalability on a broader dataset of 400 samples. Notably, MO_SCPLOWETAC_SCPLOWMO_SCPLOWUSEC_SCPLOW avoids data hallucination by defaulting to conservative false negatives when evidence is ambiguous, thereby preserving strict data integrity. By providing a fully auditable and context-aware curation pipeline, MO_SCPLOWETAC_SCPLOWMO_SCPLOWUSEC_SCPLOW offers a scalable solution for enriching public data repositories and accelerating reproducible, data-driven scientific discovery.

genomics↗

bioRxiv2025

GeneJEPA: A Predictive World Model of the Transcriptome

We introduce GO_SCPLOWENEC_SCPLOWJO_SCPLOWEPAC_SCPLOW, a self-supervised foundation model that learns a predictive world model of single-cell transcriptomes. Based on the Joint-Embedding Predictive Architecture, GO_SCPLOWENEC_SCPLOWJO_SCPLOWEPAC_SCPLOW predicts latent representations of masked gene sets from visible context, a shift away from reconstructing noisy expression values and toward world-model style inference over cellular state. To realize this at scale, a Perceiver encoder handles variable gene sets at fixed cost, and a tokenizer jointly represents gene identity and continuous expression using Fourier features. Trained on the Tahoe-100M atlas, GO_SCPLOWENEC_SCPLOWJO_SCPLOWEPAC_SCPLOW learns general representations that transfer across tissues and datasets. On downstream tasks, including drug response and perturbation prediction, it surpasses strong baselines and enables test-time scaling by progressively enlarging the cross-attention over the gene set, trading a small read cost for higher accuracy at inference. GO_SCPLOWENEC_SCPLOWJO_SCPLOWEPAC_SCPLOW moves toward foundation models that reason over gene-gene relations, enabling applications in annotation, prediction, and in-silico discovery.

genomics↗

bioRxiv2025

REVIVE-Flow: A Foundation Model for Blood DNAm Aging

Epigenetic clocks can predict biological age but cannot prescribe the interventions needed to reverse it. Here, we introduce REjuVenatIon Via Epigenetic Flow (Revive-Flow), a flow-matching model trained on a broad compendium of epigenetic blood studies to transport methylomes forward and backward in time. First, we learn the continuous vector field of aging as an Ordinary Differential Equation (ODE) within a stable, low-dimensional linear space. Then, the learned ODEs dynamics are integrated backward in time to define a natural, biologically-plausible rejuvenation trajectory. This path serves as a guide for a convex optimization problem that identifies the minimal, targeted CpG-level perturbation required to rejuvenate a sample. On a completely unseen test set comprising over 800 individuals from the European Prospective Investigation into Cancer and Nutrition (EPIC-Italy) cohort, Revive achieves 0.4 years rejuvenated per commanded year (R2 {approx} 0.99), with a smooth sparsity-effect trade-off. Extensive validation confirms the effect preserves inferred cell-type composition and targets biologically plausible loci enriched in genomic regions and pathways central to aging biology.

genomics↗

bioRxiv2025

Guided multi-agent AI invents highly accurate, uncertainty-aware transcriptomic aging clocks

Scientific discovery has long relied on human creativity, with computation limited to analysis. Here we report an AI-guided system, K-Dense, that accelerates hypothesis testing and delivers robust scientific discoveries. Trained on ARCHS4 (57,584 samples, 28 tissues, 1,039 cohorts, ages 1-114), the unified ensemble clock achieved R2 = 0.854 and MAE = 4.26 years--while uniquely providing calibrated confidence intervals. This self-aware design flags predictions made at transitional or extreme ages, where biological heterogeneity peaks, suggesting clinical utility for uncertainty itself. Development revealed stage-specific markers, including CDKN2A/p16 (senescence), AMPD3 (muscle wasting), MIR29B2CHG (progeroid traits), and SEPTIN3 (neurodegeneration resistance). Sliding-window analysis across 85 overlapping ranges uncovered wave-like shifts in gene importance, showing that transcriptomic aging signatures evolve continuously, not discretely. By transforming biological age assessment from static point estimates to calibrated predictions with explicit uncertainty, this approach establishes reliable and interpretable clocks. Beyond the clock itself, K-Dense demonstrates how guided AI can compress months of exploration into weeks, pointing toward a scalable framework for accelerated scientific discovery.

systems biology↗

bioRxiv2025

The Limitations of TabPFN for High-Dimensional RNA-seq Analysis

Tabular Prior-Data Fitted Networks (TabPFN) demonstrate remarkable performance on small-to-medium tabular datasets through in-context learning, but struggle with high-dimensional genomic data such as RNA-seq with tens of thousands of features. We investigate multiple approaches to adapt TabPFN for transcriptomic analysis using two benchmark datasets: Age-ARCHS4, a regression dataset derived from the ARCHS4 dataset (57,873 samples, 10,000 genes), and an Inflammatory Bowel Disease (IBD) classification dataset encompassing Crohns Disease and Ulcerative Colitis samples (2,490 samples, 10,000 genes). Our experimental design proceeds in two phases: first evaluating existing optimization methods, then testing novel adaptations including (1) self-supervised embedding learning and (2) Bulk-Former integration. We demonstrate that when constrained to equal training conditions (500 features, 10,000 samples), TabPFN outperforms classical baselines like random forest and XGBoost. However, when classical methods utilize full feature sets while TabPFN adaptations attempt to handle higher-dimensional data, all TabPFN variants consistently underperform the naive baseline. Our findings reveal fundamental limitations in current approaches to adapting TabPFN for genomic applications, showing that architectural modifications paradoxically degrade performance, while intelligent metadata-based subgrouping emerges as the most effective strategy for deploying TabPFN on biological data.

bioinformatics↗

bioRxiv2024

Human Mesofluidic Intestinal Model for Studying Transport of Drug Carriers and Bacteria Through a Live Mucosal Barrier

The intestinal mucosal barrier forms a critical interface between lumen contents such as bacteria, drugs, and drug carriers and the underlying tissue. Current in vitro intestinal models, while recapitulating certain aspects of this barrier, generally present challenges with respect to imaging transport across mucus and uptake into enterocytes. A human mesofluidic small intestinal chip was designed to enable facile visualization of a mucosal interface created by growing primary human intestinal cells on a vertical hydrogel wall separating channels representing the intestinal lumen and circulatory flow. Type I collagen, fortified via cross-linking to prevent deformation and leaking during culture, was identified as a suitable gel wall material for supporting primary organoid-derived human duodenal epithelial cell attachment and monolayer formation. Addition of DAPT and PGE2 to culture medium paired with air-liquid interface culture increased the thickness of the mucus layer on epithelium grown within the device for 5 days from approximately 5 mm to 50 m, making the model suitable for revealing intriguing features of interactions between luminal contents and the mucus barrier using live cell imaging. Time-lapse imaging of nanoparticle diffusion within mucus revealed a zone adjacent to the epithelium largely devoid of nanoparticles up to 4.5 hr after introduction to the lumen channel, as well as pockets of dimly lectin-stained mucus within which particles freely diffused, and apparent clumping of particles by mucus components. Multiple particle tracking conducted on the intact mucus layer in the chip revealed significant size-dependent differences in measured diffusion coefficients. E. coli introduced to the lumen channel were freely mobile within the mucus layer and appeared to intermittently contact the epithelial surface over 30 minute periods of culture. Mucus shedding into the lumen and turnover of mucus components within cells were visualized. Taken together, this system represents a powerful tool for visualization of interactions between luminal contents and an intact live mucosal barrier.

bioengineering↗

bioRxiv2024

A Strategic Blend of Stabilizing Polymers to Control Particle Surface Charge for Enhanced Mucus Transport and Cell Binding

Mucus layers, viscoelastic gels abundant in anionic mucin glycoproteins, obstruct therapeutic delivery across all mucosal surfaces. We found that strongly positively charged nanoparticles (NPs) rapidly adsorb a mucin protein corona in mucus, impeding cell binding and uptake. To overcome this, we developed mucus-evading, cell-adhesive (MECS) NPs with variable surface charge using Flash NanoPrecipitation, by blending a neutral poly(ethylene glycol) (PEG) corona for mucus transport with a small amount, 5 wt%, of polycationic dimethylaminoethyl methacrylate (PDMAEMA) for increased cell targeting. In vitro experiments confirmed rapid mucus penetration and binding to epithelial cells by MECS NPs, suggesting a breakthrough in mucosal drug delivery.

biophysics↗

bioRxiv2024

A scalable approach to absolute quantitation in metabolomics

Mass spectrometry-based metabolomics allows for the quantitation of metabolite levels in diverse biological samples. The traditional method of converting peak areas to absolute concentrations involves the use of matched heavy isotopologues. However, this approach is laborious and limited to a small number of metabolites. We addressed these limitations by developing PyxisTM, a machine learning-based technology which converts raw mass spectrometry data to absolute concentration measurements without the need for per-analyte standards. Here, we demonstrate Pyxis performance by quantifying metabolome concentration dynamics in murine blood plasma. Pyxis performed equivalently to traditional quantitation workflows used by research institutions, with a fraction of the time needed for analysis. We show that absolute quantitation by Pyxis can be expanded to include concentrations for additional metabolites, without the need to acquire new data. Furthermore, Pyxis allows for absolute quantitation as part of an untargeted metabolomics workflow. By removing the bottleneck of per-analyte standards, Pyxis allows for absolute quantitation in metabolomics that is scalable to large numbers of metabolites. The ability of Pyxis to make concentration-based measurements across the metabolome has the potential to deepen our understanding of diverse metabolic perturbations.

systems biology↗

bioRxiv2019

Niche-inspired synthetic matrices for epithelial organoid culture

Epithelial organoids are now an important tool in fields ranging from regenerative medicine to drug discovery. Organoid culture requires Matrigel, a complex, tumor-derived, extracellular matrix. An alternative completely synthetic matrix could improve culture reproducibility, clarify mechanistic phenomena, and enable applications involving human implantation. Here, we designed synthetic matrices with tunable biomolecular and biophysical properties that allowed us to identify critical gel parameters in organoid formation. Inspired by known epithelial integrin expression in the proliferative niche of the human intestine, we identified an 2{beta}1 integrin-binding peptide as a critical component of the synthetic matrix that supports human duodenal colon and endometrial organoid propagation. We show that organoids emerge from single cells, retain their proliferative capacity, are functionally responsive to basolateral stimulation and have correct apicobasal polarity upon induction of differentiation. The local biophysical presentation of the cues, rather than bulk mechanical properties, appears to be the dominant parameter governing epithelial cell proliferation and organoid formation in the synthetic matrix.

bioengineering↗

bioRxiv2018

OrgaQuant: Intestinal Organoid Localization and Quantification Using Deep Convolutional Neural Networks

Organoid cultures are proving to be powerful in vitro models that closely mimic the cellular constituents of their native tissue. The organoids are typically expanded and cultured in a 3D environment using either naturally derived or synthetic extracellular matrices. Assessing the morphology and growth characteristics of these cultures has been difficult due to the many imaging artifacts that accompany the corresponding images. Unlike single cell cultures, there are no reliable segmentation techniques that allow for the localization and quantification of organoids in their 3D culture environment. Here we describe OrgaQuant, a deep convolutional neural network implementation that can locate and quantify the size distribution of intestinal organoids in brightfield images. OrgaQuant is an end-to-end trained neural network that requires no parameter tweaking, thus it can be fully automated to analyze thousands of images with no user intervention. To develop OrgaQuant we created a unique dataset of manually annotated intestinal organoid images and trained an object detection pipeline using TensorFlow. We have made the dataset, trained model and inference scripts publically available along with detailed usage instructions.

bioengineering↗

bioRxiv2017

Closed-loop feedback control for microfluidic systems through automated capacitive fluid height sensing

Precise fluid height sensing in open-channel microfluidics has long been a desirable feature for a wide range of applications. However, performing accurate measurements of the fluid level in small-scale reservoirs (<1mL) has proven to be an elusive goal, especially if direct fluid-sensor contact needs to be avoided. In particular, gravity-driven systems used in several microfluidic applications to establish pressure gradients and impose flow remain open-loop and largely unmonitored due to these sensing limitations. Here we present an optimized self-shielded coplanar capacitive sensor design and automated control system to provide submillimeter fluid-height resolution (~250 m) and control of small-scale open reservoirs without the need for direct fluid contact. Results from testing and validation of our optimized sensor and system also suggest that accurate fluid height information can be used to robustly characterize, calibrate and dynamically control a range of microfluidic systems with complex pumping mechanisms, even in cell culture conditions. Capacitive sensing technology provides a scalable and cost-effective way to enable continuous monitoring and closed-loop feedback control of fluid volumes in small-scale gravity-dominated wells in a variety of microfluidic applications.

bioengineering↗

bioRxiv2017

MICCS: A Fully Programmable Multipurpose Integrated Cell Culture System

With the rise of research utilizing microphysiological systems (MPSs), the need for tools that enable the physiological mimicking of the relevant cellular environment is vital. The limited ability to reproduce crucial features of the microenvironment, such as surrounding fluid flow and dynamic changes in biochemical stimuli, severely limits the types of experiments that can be carried out. Current equipment to achieve this, such as syringe and peristaltic pumps, is expensive, large, difficult to program and has limited potential for scalability. Here, we present a new pumping platform that is open-source, low-cost, modular, scalable, fully-programmable and easy to assemble that can be incorporated into cell culture systems to better recapitulate physiological environments. By controlling two commercially available piezoelectric pumps using a Raspberry Pi Zero microcontroller, the system is capable of producing arbitrary dynamic flow profiles with reliable flow rates ranging from 1 to 3,000 {micro}L/min as specified by an easily programmable Python-based script. We validated the accuracy of the flow rates, the use of time-varying profiles, and the practicality of the system by creating repeatable dynamic concentration profiles using a 3D-printed static micromixer.

bioengineering↗