Biology subjects

Joshi, C. K.

Publications and source records attributed to Joshi, C. K..

5 recordsLinked to original sources

bioRxiv2026

De novo design of RNA pseudoknots with deep learning

RNA design has been hindered by the limited accuracy of 3D structure prediction. Here, we show that intricate RNA structures can be generated with current deep learning tools through accurate de novo design of pseudoknot secondary structures. In an Eterna competition involving 57 pseudoknots, generative AI methods matched experienced human designers in solving most blind challenges, evaluated by single-nucleotide-resolution chemical mapping, compensatory mutagenesis, and cryogenic electron microscopy. AI-generated molecules with accurate secondary structures formed well-ordered 3D folds stabilized by noncanonical tertiary interactions not modeled during design. Success was guided by an RNet foundation model trained on prior chemical mapping data, suggesting that some difficult RNA design tasks may be tractable without first solving RNA 3D structure prediction.

biophysics↗

bioRxiv2025

Template-based RNA structure prediction advanced through a blind code competition

Automatically predicting RNA 3D structure from sequence remains an unsolved challenge in biology and biotechnology. Here, we describe a Kaggle code competition engaging over 1700 teams and 43 previously unreleased structures to tackle this challenge. The top three submitted algorithms achieved scores within statistical error of the winners of the recent CASP16 competition. Unexpectedly, the top Kaggle strategy involved a pipeline for discovering 3D templates, without the use of deep learning. We integrated this template-modeling pipeline and other Kaggle strategies to develop a single model RNAPro that retrospectively outperformed individual Kaggle models on the same test set. These results suggest a growing importance of template-based modeling in RNA structure prediction.

biophysics↗

bioRxiv2025

Generative inverse design of RNA structure and function with gRNAde

The design of RNA molecules with bespoke three-dimensional structures and functions is a central goal in synthetic biology and biotechnology. However, progress has been limited by the challenges of designing complex tertiary interactions such as pseudoknots, as well as engineering catalytic functions--problems that have remained largely intractable for automated methods. Here we present a high-throughput generative AI pipeline for inverse design of RNA structure and function. Central to the pipeline is gRNAde, an RNA language model conditioned on 3D backbone structures and sequence constraints. We have validated the gRNAde pipeline in a community-wide, blinded RNA design competition on the Eterna platform, where it proved able to design complex pseudoknotted RNAs at success rates matching that of human experts (95%), while significantly outperforming other physics- and AI-based automated algorithms (70%). We further demonstrate gRNAdes capabilities by generatively designing functional RNA polymerase ribozymes (RPR) with nearly 20% sequence divergence from the wild type RPR, discovering highly active variants at mutational distances inaccessible to rational design or adaptive walks by directed evolution. gRNAde thus provides an experimentally validated, open-source platform for automated design of complex RNA structures and accelerated engineering of complex RNA functions, providing a step towards programmable RNA catalysts and nanostructures. Open-source code: github.com/chaitjo/geometric-rna-design

synthetic biology↗

bioRxiv2024

gRNAde: Geometric Deep Learning for 3D RNA inverse design

Computational RNA design tasks are often posed as inverse problems, where sequences are designed based on adopting a single desired secondary structure without considering 3D conformational diversity. We introduce gRNAde, a geometric RNA design pipeline operating on 3D RNA backbones to design sequences that explicitly account for structure and dynamics. gRNAde uses a multi-state Graph Neural Network and autoregressive decoding to generates candidate RNA sequences conditioned on one or more 3D backbone structures where the identities of the bases are unknown. On a single-state fixed backbone re-design benchmark of 14 RNA structures from the PDB identified by Das et al. (2010), gRNAde obtains higher native sequence recovery rates (56% on average) compared to Rosetta (45% on average), taking under a second to produce designs compared to the reported hours for Rosetta. We further demonstrate the utility of gRNAde on a new benchmark of multi-state design for structurally flexible RNAs, as well as zero-shot ranking of mutational fitness landscapes in a retrospective analysis of a recent ribozyme. Experimental wet lab validation on 10 different structured RNA backbones finds that gRNAde has a success rate of 50% at designing pseudoknotted RNA structures, a significant advance over 35% for Rosetta. Open source code and tutorials are available at: github.com/chaitjo/geometric-rna-design

bioinformatics↗

bioRxiv2022

Hypergraph factorisation for multi-tissue gene expression imputation

AO_SCPLOWBSTRACTC_SCPLOWIntegrating gene expression across scales and tissues is crucial for understanding the biological mechanisms that drive disease and characterise homeostasis. However, traditional multi-tissue integration methods cannot handle uncollected tissues or rely on genotype information, which is subject to privacy concerns and often unavailable. To address these challenges, we present HYFA (Hypergraph Factorisation), a novel method for joint imputation of multi-tissue and cell-type gene expression. HYFA imputes tissue-specific gene expression via a specialised graph neural network operating on a hypergraph of individuals, metagenes, and tissues. HYFA is genotype- agnostic, supports a variable number of collected tissues per individual, and imposes strong inductive biases to leverage the shared regulatory architecture of tissues. In performance comparison on data from the Genotype Tissue Expression project, HYFA achieves superior performance over existing transcriptome imputation methods, especially when multiple reference tissues are available. Through transfer learning on a paired single-nucleus RNA-seq (snRNA-seq) dataset, we further show that HYFA can accurately resolve cell-type signatures from bulk gene expression, highlighting the methods ability to leverage gene expression programs underlying cell-type identity, even in tissues that were never observed in the training set. Using Gene Set Enrichment Analysis, we find that the metagenes learned by HYFA capture information about known biological pathways. Notably, the HYFA-imputed dataset can be used to identify regulatory genetic variations (eQTLs), with substantial gains over the original incomplete dataset. Our framework can accelerate effective and scalable integration of tissue and cell-type gene expression biorepositories.

bioinformatics↗