Biology subjects

Cristea, S.

Publications and source records attributed to Cristea, S..

5 recordsLinked to original sources

bioRxiv2026

evoCancerGPT: Generating Zero-Shot Single-Cell and Single-Sample Cancer Progression Through Transfer Learning

Cancer evolution is driven by complex changes in gene expression as cells transition and change states during tumorigenesis. Single-cell RNA sequencing has provided snapshot insights into how the transcriptomics of tumors evolve, but whether the existing knowledge can be used to reliably learn and generate the patterns behind the evolution of cancers remains unknown. Here, we introduce evoCancerGPT, a generative pre-trained transformer decoder-only single-cell foundation model designed to forecast future gene expression profiles in cancer evolution by leveraging previous cell states at the level of single patients. This model integrates the continuous gene expression data of each cell to create a comprehensive representation of a cell token. Training sentences are constructed for each cancer type, each patient and each cell type separately, ordered via inferred pseudotime algorithms, using 2.76 million cell tokens, each with 12,639 genes, spanning 7 cancer types. By learning from long-range dependencies between cells arranged in pseudotime from a large corpus of data, evoCancerGPT captures key transitions in cancer evolution, achieving high concordance to ground truth trajectories and outperforming linear and scGPT baselines in held-out test samples in low-context scenarios. Our work suggests evoCancerGPTs potential utility in characterizing tumor progression at a single-cell and single-patient level and ultimately contributing to more personalized cancer care.

bioinformatics↗

bioRxiv2025

Biological Reasoning with Reinforcement Learning through Natural Language Enables Generalizable Zero-Shot Cell Type Annotations

Single-cell RNA-sequencing (scRNAseq) has reshaped biomedical research, enabling the high-resolution characterization of cellular populations. Yet cell type annotation, a process typically performed by domain experts interpreting gene expression patterns by manual curation or with specialized algorithms, remains labor-intensive and limited by prior knowledge. In addition, while reasoning large language models (LLMs) have demonstrated remarkable performance on mathematics, coding and general-reasoning benchmarks, their potential in scRNAseq analyses remains underexplored. Here, we investigate the advantages and limitations of employing DeepSeek-R1-0528, a recently developed open-source 671B-parameter reasoning LLM, for zero-shot scRNAseq cell type annotation. We find that DeepSeek-R1 prompted with a ranked list of 10 differentially expressed marker genes per cluster of single cells outperforms both its reasoning-enhanced, non-reasoning equivalent (DeepSeek-V3-0324) and GPT-4o in cluster-level annotations. At the level of single cells, DeepSeek-R1 prompted with the top 500 expressed genes in a cell outperforms its non-reasoning counterpart DeepSeek-V3, illustrating test-time scaling for bioinformatics tasks through natural language. Running DeepSeek-R1 in zero-shot classifier mode, with a prompt that presents a broad catalogue of cell type labels to choose from, improves its performance and generalizability across different datasets. On data curated by the expert model scTab (termed in-domain data), the DeepSeek-R1 classifiers perform better than the expert model scGPT and on par with the specialized cell genomics LLM C2S-Scale-1B, but lag behind scTab. On out-of-distribution data unseen by the two expert models, DeepSeek-R1 and its classifier versions generalize better and outperform the other models in the majority of the evaluated datasets. Notably, DeepSeek-R1 supports its cell type calls with interpretable textual biological rationales underlying its reasoning, providing a learning opportunity for researchers. Nevertheless, peak annotation performance remains modest, highlighting the intrinsic complexity of scRNAseq cell type annotation.

bioinformatics↗

bioRxiv2022

A paracrine circuit of IL-1b/IL-1R1 between myeloid and tumor cells drives glioblastoma progression

Monocytes and monocyte-derived macrophages (MDM) from blood circulation infiltrate and promote glioblastoma growth. Here we discover that glioma cells induce the expression of potent pro-inflammatory cytokine IL-1{beta} in MDM, which engages IL-1R1 in glioma cells, activates NF-{kappa}B pathway, and subsequently leads to the induction of monocyte chemoattractant proteins (MCPs). Thus, a feedforward paracrine circuit of IL-1{beta}/IL-1R1 between the tumors and MDM creates an interdependence driving glioblastoma progression. Locally antagonizing IL-1{beta}/IL-1R1 leads to reduced MDM infiltration, diminished tumor growth, reduced exhausted CD8+ T cells, and thereby extends the survival of tumor-bearing mice. In contrast to IL-1{beta}, IL-1a exhibits anti-tumor effects. Genetic deletion of Il1a is associated with decreased recruitment of lymphoid cells and loss of interferon (IFN) signaling in various immune populations and subsets of malignant cells. IL-1{beta} antagonism of IL-1{beta} should be considered as an effective anti-glioblastoma therapy. Graphical Abstract O_FIG O_LINKSMALLFIG WIDTH=196 HEIGHT=200 SRC="FIGDIR/small/486888v1_ufig1.gif" ALT="Figure 1"> View larger version (66K): org.highwire.dtl.DTLVardef@1c99e57org.highwire.dtl.DTLVardef@1ba9d24org.highwire.dtl.DTLVardef@a2c418org.highwire.dtl.DTLVardef@1f8091c_HPS_FORMAT_FIGEXP M_FIG C_FIG

cancer biology↗

bioRxiv2018

Cyclin D-Cdk4,6 drives cell cycle progression via the retinoblastoma protein’s C-terminal helix

The cyclin-dependent kinases Cdk4 and Cdk6 form complexes with D-type cyclins to drive cell proliferation. A well-known target of cyclin D-Cdk4,6 is the retinoblastoma protein, Rb, which inhibits cell cycle progression until its inactivation by phosphorylation. However, the role of cyclin D-Cdk4,6 phosphorylation of Rb in cell cycle progression is unclear because Rb can be phosphorylated by other cyclin-Cdk complexes and cyclin D-Cdk4,6 complexes have other targets that may drive cell division. Here, we show that cyclin D-Cdk4,6 docks one side of an alpha-helix in the C-terminus of Rb, which is not recognized by cyclins E, A, and B. This helix-based docking mechanism is shared by the p107 and p130 Rb-family members across metazoans. Mutation of the Rb C-terminal helix prevents phosphorylation, promotes G1 arrest, and enhances Rbs tumor suppressive function. Our work conclusively demonstrates that the cyclin D-Rb interaction drives cell division and defines a new class of cyclin-based docking mechanisms.

cell biology↗

bioRxiv2018

ModulOmics: Integrating Multi-Omics Data to Identify Cancer Driver Modules

The identification of molecular pathways driving cancer progression is a fundamental unsolved problem in tumorigenesis, which can substantially further our understanding of cancer mechanisms and inform the development of targeted therapies. Most current approaches to address this problem use primarily somatic mutations, not fully exploiting additional layers of biological information. Here, we describe ModulOmics, a method to de novo identify cancer driver pathways, or modules, by integrating multiple data types (protein-protein interactions, mutual exclusivity of mutations or copy number alterations, transcriptional co-regulation, and RNA co-expression) into a single probabilistic model. To efficiently search the exponential space of candidate modules, ModulOmics employs a two-step optimization procedure that combines integer linear programming with stochastic search. Across several cancer types, ModulOmics identifies highly functionally connected modules enriched with cancer driver genes, outperforming state-of-the-art methods. For breast cancer subtypes, the inferred modules recapitulate known molecular mechanisms and suggest novel subtype-specific functionalities. These findings are supported by an independent patient cohort, as well as independent proteomic and phosphoproteomic datasets.

cancer biology↗