AI Virtual Cell. GREmLN: A Cellular Regulatory Network-Aware Transcriptomics Foundation Model.

AI Virtual Cell. Research. Gen AI. LLMs.  AI in BioTech. Part-9

AI Virtual Cell. GREmLN: A Cellular Regulatory Network-Aware
Transcriptomics Foundation Model.

Date: 7/13/2025

Fields: Gen AI, Biology, Genomics, DNA, RNA, BioTech, LLMs, VLMs.

AI Virtual Cell. Research Timeline

Listen to Podcast

Related Research

Paper: GREmLN: A Cellular Regulatory Network-Aware Transcriptomics Foundation Model. Mingxuan Zhang, Vinay Swamy, Rowan Cassius, Léo Dupire, Charilaos Kanatsoulis, Evan Paull, Mohammed AlQuraishi, Theofanis Karaletsos, Andrea Califano. DOI: https://doi.org/10.1101/2025.07.03.663009

Research Timeline

2003

Bar-Joseph et al. publication: Computational discovery of gene modules and regulatory networks

Published, contributing to the understanding of gene regulatory networks.

2006

Margolin and Nemenman publish ARACNe

The ARACNe algorithm is introduced as a method for reconstructing gene regulatory networks, which becomes a key component in later foundation models.

2009

Hecker et al. publication: Gene regulatory network inference: data integration in dynamic models—a review

Provides an overview of GRN inference, highlighting the importance of data integration.

2013

Paull et al. publish TIE-DIE

Discovering causal pathways linking genomic events to transcriptional states using tied diffusion through interacting events (tiedie) is published, demonstrating early uses of diffusion models in biological data.

2014

Chen, Alvarez, Talos et al. publication: Identification of causal genetic drivers of human disease through systems-level analysis of regulatory networks

Published, highlighting the use of regulatory networks in understanding disease.

2015

Leiserson et al. publication: Pan-cancer network analysis identifies combinations of rare somatic mutations across pathways and protein complexes

Further demonstrates the utility of network analysis in cancer.

2016

Alvarez et al. publish on network-based inference of protein activity

Network-based inference of protein activity helps functionalize the genetic landscape of cancer and Functional characterization of somatic mutations in cancer using network-based inference of protein activity are published, demonstrating the biological meaningfulness of ARACNe-inferred networks.

2017

Goldsborough et al. publish CytoGAN

CytoGAN: Generative Modeling of Cell Images introduces Generative Adversarial Networks for morphological profiling of cell images, enabling visualization of cellular variations.

Califano and Alvarez publish on tumor architecture

The recurrent architecture of tumour initiation, progression and drug sensitivity is published, emphasizing the role of gene regulatory networks in disease.

2018

Simm et al. publish on repurposing image assays

Repurposing High-Throughput Image Assays Enables Biological Activity Prediction for Drug Discovery demonstrates the ability to predict compound activity from images, boosting hit rates in drug discovery.

2019

Lafarge et al. publish on VAE for cell imaging

Capturing Single-Cell Phenotypic Variation via Unsupervised Representation Learning introduces a VAE framework for cell image representation learning, improving classification accuracy over GANs.

Paszke, Gross, Massa, et al. publish PyTorch

PyTorch, a deep learning library, is released, which becomes a foundational tool for implementing models like GREmLN.

2020

Dwivedi and Bresson publish on Graph Transformers

A generalization of transformer networks to graphs contributes to the growing field of integrating graph structural priors into Transformer architectures.

Mercatelli et al. publication: Gene regulatory network inference resources: A practical overview

Reviews methods for GRN inference.

Seyone Chithrananda and Ramsundar publish ChemBERTa

Chemberta: Large-scale self-supervised pretraining for molecular property prediction focuses on small molecule foundation models.

Zhang et al. publish Graph-BERT

Graph-bert: Only attention is needed for learning graph representations further explores graph representations in the context of Transformers.

2021

Way et al. publish on predicting cell health phenotypes

Predicting cell health phenotypes using image-based morphology profiling shows that Cell Painting images can predict cell health readouts.

Rives et al. publish ESM

Biological structure and function emerge from scaling unsupervised learning to 250 million protein sequences introduces ESM, an early biological foundation model focused on protein sequences.

Paull, Aytes, et al. publication

A modular master regulator landscape controls cancer transcriptional identity uses ARACNe-derived networks in cancer research.

Wu and Jain et al. publication

Representing long-range context for graph neural networks with global attention contributes to methods for global attention in graph neural networks.

2022

Chow et al. publish on predicting drug polypharmacology

Predicting drug polypharmacology from cell morphology readouts using variational autoencoder latent space arithmetic uses VAEs to predict drug polypharmacology from cell morphology and gene expression data.

Caicedo et al. publish on Cell Painting predicting cancer variants

Cell Painting predicts impact of lung cancer variants demonstrates the use of Cell Painting for assessing the functional impact of gene variants in lung cancer.

Haghighi et al. publish high-dimensional profiles

High-dimensional gene expression and morphology profiles of cells across 28,000 genetic and chemical perturbations provides datasets for developing multimodal methodologies in drug discovery.

Theodoris and Xiao et al. publish Geneformer

Transfer learning enables predictions in network biology introduces Geneformer, a Transformer-based scRNA-seq foundation model.

Tri, Fu, Ermon, et al. publish FlashAttention

Flashattention: Fast and memory-efficient exact attention with io-awareness is published, a technique used in GREmLN for efficiency.

Moshkov et al. publish on predicting compound activity

Predicting compound activity from phenotypic profiles and chemical structures evaluates the predictive power of chemical structures, cell morphology, and transcriptional profiles for assay outcomes.

2023

Wakui et al. publish on predicting gene expression from morphology

Predicting reprogramming-related gene expression from cell morphology in human induced pluripotent stem cells explores the relationship between gene expression and morphology in iPSCs.

Zeleke, Pan, et al. publication

Network-based assessment of hdac6 activity predicts preclinical and clinical responses to the hdac6 inhibitor ricolinostat in breast cancer further validates the use of networks in clinical contexts.

Grunn A et al. Mundi PS, Dela Cruz FS. publication

A transcriptome-based precision oncology platform for patient-therapy alignment in a diverse set of treatment-resistant malignancies. uses ARACNe in clinical trials.

2024

Cui and Wang et al. publish scGPT

scgpt: Toward building a foundation model for single-cell multi-omics using generative ai introduces scGPT, a Transformer-based scRNA-seq foundation model.

Hao and Gong et al. publish scFoundation

Large-scale foundation model on single-cell transcriptomics introduces scFoundation, another Transformer-based scRNA-seq foundation model.

Zou, Tao, et al. publish on RNA foundation model

A large-scale foundation model for rna function and structure prediction contributes to RNA sequence foundation models.

2025

July

3

Zhang, Swamy, et al. publish GREmLN preprint

GREmLN: A Cellular Regulatory Network-Aware Transcriptomics Foundation Model is published as a preprint, introducing a new scRNA foundation model that integrates gene regulatory networks into its attention mechanism. This model aims to overcome limitations of previous Transformer-based models by explicitly encoding gene-gene relationships.