Juho Rousu

Domain-adversarial learning predicts clinically actionable drug combination synergy in leukemia patients using bulk transcriptomics data

This paper develops a domain-adversarial neural network to predict drug combination synergy in leukemia patients using bulk transcriptomics data, showing improved accuracy over existing methods.

Structured Chemical Reaction Modeling with Multitask Graph Neural Networks

This paper introduces a multitask graph neural network for chemical reaction modeling that jointly predicts atom mappings, reaction centers, and classifications, achieving state-of-the-art results on a benchmark.

Scaling up drug combination surface prediction

This paper develops an improved machine learning model for predicting drug combination dose-response surfaces, showing better performance than scalar-valued synergy score methods.

Predicting drug combination response surfaces

This paper introduces comboKR, a structured prediction method for drug combination response surfaces, showing improved performance in predicting responses for new drugs.

Attention-based approach to predict drug-target interactions across seven target superfamilies

This paper introduces an attention-based method to predict drug-target interactions across seven superfamilies, achieving high correlation and identifying missing activities in approved drugs.

Joint structural annotation of small molecules using liquid chromatography retention order and tandem mass spectrometry data

LC-MS2Struct improves small molecule structural annotation by combining retention order and MS2 data with a structured prediction model, achieving higher accuracy than existing methods.

Strain design optimization using reinforcement learning

This paper uses multi-agent reinforcement learning to optimize microbial strain production, showing improved convergence and stability in synthetic biology applications.

Ranking microbial metabolomic and genomic links in the NPLinker framework using complementary scoring functions

The paper introduces NPLinker, a framework that combines scoring functions to link microbial genomic and metabolomic data, improving identification of biosynthetic gene clusters and their associated metabolites.

Modeling drug combination effects via latent tensor reconstruction

The paper introduces comboLTR, a method for predicting drug combination effects using latent tensor reconstruction, outperforming existing methods in speed and accuracy.

comboFM: leveraging multi-way interactions for systematic prediction of drug combination effects

comboFM predicts drug combination effects using machine learning, achieving high accuracy in cancer cell line screening and validating novel synergies.

Classes for the masses: Systematic classification of unknowns using fragmentation spectra

CANOPUS is a tool that predicts compound classes from fragmentation spectra, enabling annotation of unknown compounds without reference data.

Probabilistic Framework for Integration of Mass Spectrum and Retention Time Information in Small Molecule Identification

This paper introduces a probabilistic framework that integrates mass spectrometry and retention time data to improve small molecule identification in LC-MS experiments.

Principal Metabolic Flux Mode Analysis

The paper introduces PMFA, a method combining PCA and stoichiometric flux analysis for metabolic flux mode analysis, showing its versatility through experiments.

Predicting Protein Producibility In Filamentous Fungi

The paper predicts protein producibility in filamentous fungi using machine learning, finding that models trained on one species can generalize to others with limited accuracy loss.