Philippe Nghe

DORA predicts dose-dependent changes in transcriptomes and cell viability using a deep learning autoencoder, outperforming existing methods in viability prediction.

This paper develops a microfluidic platform for 3D tumor-stroma co-cultures that stabilizes ECM and supports long-term drug testing with high-content imaging and AI analysis.

The paper uses probabilistic generative models to estimate the number of self-reproducing RNAs and finds that at least 10^39 such ribozymes may exist.

The paper improves generative models for biological sequences by integrating experimental feedback, increasing functional sequence generation from 6.7% to 63.7% in ribozyme design.

This paper explores active learning strategies for synergistic drug discovery, finding that environmental features improve predictions and that small batch sizes with dynamic strategies can identify 60% of synergistic pairs from 10% of the combinatorial space.

This paper introduces D-LIM, an interpretable neural network for genotype-to-fitness mapping, achieving state-of-the-art accuracy in predicting gene and gene-environment interactions in mutational scanning data.

This paper introduces eaDCA, a sparse generative model for RNA families that efficiently predicts functional sequences and mutation effects, showing comparable performance to complex methods with fewer parameters.

The paper shows global regulators enable adaptation to a growth-motility trade-off in E. coli by modulating both phenotypes, while local regulators mainly affect motility.

The paper demonstrates metabolic coupling during rapid adaptation of single yeast cells by measuring metabolism and division in thousands of individual cells.

The paper quantifies fitness interactions between expression variants of two metabolic genes across environments and shows these can be explained by a model incorporating catabolic flux, metabolite toxicity, and expression costs.