Toward a foundation model of causal cell and tissue biology with a Perturbation Cell and Tissue Atlas

Cell, Год журнала: 2024, Номер 187(17), С. 4520 - 4545

Опубликована: Авг. 1, 2024

Язык: Английский

---

Deep learning-based predictions of gene perturbation effects do not yet outperform simple linear methods

bioRxiv (Cold Spring Harbor Laboratory), Год журнала: 2024, Номер unknown

Опубликована: Сен. 19, 2024

Abstract Advanced deep-learning methods, such as foundation models, promise to learn representations of biology that can be employed predict in silico the outcome unseen experiments, effect genetic perturbations on transcriptomes human cells. To see whether current models already reach this goal, we benchmarked five and two other deep learning against deliberately simplistic linear baselines. For combinatorial genes for which only individual single had been seen, find learning-based approaches did not perform better than a simple additive model. yet outper-form baseline predicting mean across training perturbations. We hypothesize poor performance is partially because pre-training data observational; show model reliably outperforms all when pre-trained another perturbation dataset. While neural networks representation biological systems prediction experimental outcomes plausible, our work highlights need clear setting objectives critical benchmarking direct research efforts. Contact [email protected]

Язык: Английский

---

AI-empowered perturbation proteomics for complex biological systems

Cell Genomics, Год журнала: 2024, Номер 4(11), С. 100691 - 100691

Опубликована: Ноя. 1, 2024

SummaryThe insufficient availability of comprehensive protein-level perturbation data is impeding the widespread adoption systems biology. In this perspective, we introduce rationale, essentiality, and practicality proteomics. Biological are perturbed with diverse biological, chemical, and/or physical factors, followed by proteomic measurements at various levels, including changes in protein expression turnover, post-translational modifications, interactions, transport, localization, along phenotypic data. Computational models, employing traditional machine learning or deep learning, identify predict responses, mechanisms action, functions, aiding therapy selection, compound design, efficient experiment design. We propose to outline a generic PMMP (perturbation, measurement, modeling prediction) pipeline build foundation models other suitable mathematical based on large-scale Finally, contrast between artificially naturally highlight importance proteomics for advancing our understanding predictive biological systems.

Язык: Английский

---

Simple controls exceed best deep learning algorithms and reveal foundation model effectiveness for predicting genetic perturbations

bioRxiv (Cold Spring Harbor Laboratory), Год журнала: 2025, Номер unknown

Опубликована: Янв. 8, 2025

Abstract Modeling genetic perturbations and their effect on the transcriptome is a key area of pharmaceutical research. Due to complexity transcriptome, there has been much excitement development in deep learning (DL) because its ability model complex relationships. In particular, transformer-based foundation paradigm emerged as gold-standard predicting post-perturbation responses. However, understanding these increasingly models evaluating practical utility lacking, along with simple but appropriate benchmarks compare predictive methods. Here, we present baseline method that outperforms both state art (SOTA) DL other proposed simpler neural architectures, setting necessary benchmark evaluate field prediction. We also elucidate for task prediction via generalizable fine-tuning experiments can be translated different applications tasks interest. Furthermore, provide corrected version popular dataset used benchmarking perturbation models. Our hope this work will properly contextualize further space control procedures.

Язык: Английский

---

Machine learning to dissect perturbations in complex cellular systems

Computational and Structural Biotechnology Journal, Год журнала: 2025, Номер 27, С. 832 - 842

Опубликована: Янв. 1, 2025

Язык: Английский

---

Immune-mediated cardiac development and regeneration

Seminars in Cell and Developmental Biology, Год журнала: 2025, Номер 171, С. 103613 - 103613

Опубликована: Май 1, 2025

The complex interplay between the immune and cardiovascular systems during development, homeostasis regeneration represents a rapidly evolving field in cardiac biology. Single cell technologies, spatial mapping computational analysis have revolutionised our understanding of diversity functional specialisation cells within heart. From earliest stages cardiogenesis, where primitive macrophages guide heart tube formation, to choreography inflammation its resolution regeneration, emerge as central orchestrators fate. Translating these fundamental insights into clinical applications major challenge opportunity for field. In this Review, we decode immunological blueprint development transform disease treatment unlock regenerative capacity human

Язык: Английский

---

Multimodal weakly supervised learning to identify disease-specific changes in single-cell atlases

bioRxiv (Cold Spring Harbor Laboratory), Год журнала: 2024, Номер unknown

Опубликована: Июль 29, 2024

Abstract Multimodal analysis of single-cell samples from healthy and diseased tissues at various stages provides a comprehensive view that identifies disease-specific cells, their molecular features aids in patient stratification. Here, we present MultiMIL, novel weakly-supervised multimodal model designed to construct references prioritize phenotype-specific cells via classification. MultiMIL effectively integrates modalities, even when they only partially overlap, providing robust representations for downstream analyses such as phenotypic prediction cell prioritization. Using multiple-instance learning approach, aggregates cell-level measurements into sample-level states through attention-based scoring. We demonstrate accurately blood lung samples, identifying disease-associated genes achieving superior classification accuracy compared existing methods. anticipate will become an essential tool querying multiomic atlases, enhancing our understanding disease mechanisms informing targeted treatments.

Язык: Английский

---

Unifying Genetic and Chemical Perturbagen Representation through a Hybrid Deep Learning Framework

bioRxiv (Cold Spring Harbor Laboratory), Год журнала: 2025, Номер unknown

Опубликована: Фев. 7, 2025

Artificial Intelligence virtual cell (AIVC) holds transformative potential for biomedical research. Central to this vision is the systematic modeling of genetic and chemical perturbation phenotypes accurately predict cellular dynamic states from diverse interventions. However, disparities in screening agents, library scales, experimental technologies, data production efficiency hinder integration, modeling, analysis cross-data. Here we present UniPert-G2CP , a two-phase deep learning approach comprising i) UniPert, multimodal molecular representation model that bridges domains, ii) G2CP (Genetic-to-Chemical Perturbation transfer learning), which systematically transforms CRISPR screen-based insights into cost-effective silico drug screening. UniPert not only encodes perturbagens unified functionally interpretable sematic embedding space, but also improves phenotypic effect prediction previously unseen gene perturbations treatments. Building upon successfully modeled large-scale post-perturbation spanning 4,994 7,821 compound perturbagens, while reducing costs by over 60%. We demonstrate enables efficient, generalizable simulations multicellular, multi-domain cause-effect spaces, revealing differential biological causality informing mechanism-driven therapy. opens new avenues causal foundation building, AIVC creation, AI-powered precision medicine.

Язык: Английский

---

GEMDiff: a diffusion workflow bridges between normal and tumor gene expression states: a breast cancer case study

Briefings in Bioinformatics, Год журнала: 2025, Номер 26(2)

Опубликована: Март 1, 2025

Abstract Breast cancer remains a significant global health challenge due to its complexity, which arises from multiple genetic and epigenetic mutations that originate in normal breast tissue. Traditional machine learning models often fall short addressing the intricate gene interactions complicate drug design treatment strategies. In contrast, our study introduces GEMDiff, novel computational workflow leveraging diffusion model bridge expression states between tumor conditions. GEMDiff augments RNAseq data simulates perturbation transformations states, enhancing biomarker identification. can handle large-scale without succumbing scalability stability issues plague other generative models. By avoiding need for task-specific hyper-parameter tuning specific loss functions, be generalized across various tasks, making it robust tool analysis. The model’s ability augment RNA-seq simulate perturbations provides valuable researchers. This capability used generate synthetic training models, thereby issue of limited biological performance predictive effectiveness is demonstrated through case using mRNA data, identifying 307 core genes involved transition state. open source available at https://github.com/xai990/GEMDiff.git under MIT license.

Язык: Английский

---

Dose-response assessment of dipentyl phthalate effects on testosterone production and morphogenesis of late-gestation fetal rat testis

Environment International, Год журнала: 2025, Номер 200, С. 109551 - 109551

Опубликована: Май 23, 2025

Dipentyl phthalate (DPeP) is a potent male reproductive toxicant that reduces fetal testicular testosterone production and induces abnormal testis morphology, including multinucleated germ cells (MNGs). We aimed to test whether production, MNG density, or gene expression would be most sensitive DPeP exposure determine which transcriptomic processes are initiated at the lowest dose. Timed pregnant Sprague Dawley rats were exposed 0, 1, 11, 33, 100, 300 mg DPeP/kg/d by oral gavage from GD 17-21. For comparison DPeP, additional vinclozolin, prochloraz, acetaminophen, mono-(2-ethylhexyl) tetrabromophthalate, dexamethasone. Testosterone was measured using an ex vivo culture assay, MNGs quantified on sections, testes used for RNA-seq, immunofluorescence, in situ hybridization. Benchmark dose (BMD) analysis compare apical endpoints expression. dose-dependently reduced increased density. ED50 density lower than but BMD10 values similar. The BMD estimates toxicity (MNGs) (R-RNO-210991: basigin interactions) 2.675 mg/kg/d 2.44 mg/kg/d, respectively. altered sets related steroidogenesis, gonad development, epithelial cell differentiation, vasculature development. conclude inhibition of induction have similar utility quantification dose-response context risk assessment. RNA-seq data suggest differentiation patterning may contribute mechanisms rat testis.

Язык: Английский

---