Image credits: ©Johan Jarnestad/The Royal Swedish Academy of Sciences

CNC-UC / CiBB scientists congratulate the Nobel Prize awardees John J.Hopfield and Geoffrey E.Hinton for foundational discoveries at the intersection of neurobiology and computation. Their research achievements have been instrumental for the development of Computational and Theoretical Neuroscience and for the inventions behind modern machine learning tools and algorithms. These discoveries are key to develop mathematical models that could incorporate the mechanisms of biological neural networks and provide principled explanations for psychological and neural data. 

At CNC-UC / CiBB, our researchers continue to push the boundaries of theoretical and computational neuroscience and neuro-inspired computing, as well as the application of machine learning in computational biology. For example: 

• POSEIDON: Researchers from the Data-Driven Molecular Design Lab and Advanced Therapies Groups developed POSEIDON as a comprehensive database for Cell-penetrating peptides (CPPs) that includes quantitative uptake values and physicochemical properties. A machine-learning model built on these data predicts CPP uptake efficiency, facilitating the development of CPP-based therapeutics (paper here).
• Synpred: Researchers from the Data-Driven Molecular Design Lab have created Synpred, an ensemble learning model that predicts the effects of drug combinations in cancer therapy. By integrating multi-omics data, Synpred identified synergistic interactions to enhance personalised treatment strategies (paper here).
• DELFOS: Researchers from the Data-Driven Molecular Design Lab developed DELFOS, a deep learning model grounded in the physics of neural networks, to predict drug sensitivity in cancer cells. DELFOS integrates multiomics data and drug structural information to achieve state-of-the-art performance in predicting cancer drug responses (paper here).
• DrugTax: Researchers from the Data-Driven Molecular Design Lab have developed DrugTax, a Python-based tool designed to categorise small molecules using SMILES strings. With over 38,000 downloads, DrugTax has enhanced explainability in AI-driven drug discovery by organising compounds into 26 superclasses, supporting large-scale analyses (paper here).
• SPOTONE: Researchers from the Data-Driven Molecular Design Lab developed SPOTONE, a new machine learning predictor that accurately classifies protein hot spots (HS) via sequence-only features, bypassing the need for three-dimensional structures. The predictor is freely available on a web server, requiring only the submission of FASTA files with protein sequences (paper here).
• SpotOn: Researchers from the Data-Driven Molecular Design Lab and Utrecht University developed SpotOn, a web server for identifying and classifying interfacial residues as hot spots (HS) and null spots (NS). SpotOn has been used by more than 20,000 users, providing a valuable tool for protein interaction studies (paper here).
• SicknessMiner: Researchers from the Data-Driven Molecular Design Lab and Advanced Computing Systems (INESC TEC) developed SicknessMiner, a biomedical text-mining tool designed to centralise disease-disease associations (DDAs) related to blood cancers. SicknessMiner employs named entity recognition and normalisation to extract DDAs from raw input data. This tool offers valuable insights into disease relationships and complements the existing resources (paper here).
• ViralBindPredict: Researchers from the Data-Driven Molecular Design Lab and the Information and Decision Support Systems (INESCID) have created ViralBindPredict, a deep learning model that predicts ligand-binding sites in viral proteins using only sequence-derived data, bypassing the need for 3D structures. This tool has accelerated antiviral drug discovery, especially for rapidly mutating viruses (paper in finalization).
• GPCR-A17 MAAP: Researchers from the Data-Driven Molecular Design Lab introduced GPCR-A17 MAAP, an ensemble machine learning model designed to predict the functional roles of ligands in G Protein-Coupled Receptors (GPCR-A17 subfamily), which are particularly relevant to neurological diseases. With a dataset of over 3,000 unique ligands and more than 6,900 protein-ligand interactions, this model achieved high classification performance and is poised to accelerate drug discovery (paper submitted).
• Computational Neuroscience: Renato Duarte, a researcher from the Neural Circuits of Social Behavior group and DYNABrain project investigates population dynamics and develops mathematical models of biological neural circuits. Strongly constrained and inspired by the developments introduced by the Nobel laureates, he uses recurrent neural network models and investigates them in functional contexts using the reservoir computing paradigm. These approaches have branched out from the pioneering work of John Hopfield on auto-associative memory and pattern completion in symmetric recurrent neural networks. Some examples of this work can be found in these three papers - here, here and here
• Behavioral analysis with deep learning: The foundational discoveries introduced by Nobel laureate Geoffrey Hinton have been instrumental to the development of modern deep learning algorithms. Researchers at the DYNABrain project regularly use these tools to annotate and analyse rat's prosocial behavior (paper here). 

As stated on the official award website “They used physics to find patterns in information:Machine learning has long been important for research, including the sorting and analysis of vast amounts of data. John Hopfield and Geoffrey Hinton used tools from physics to construct methods that helped lay the foundation for today’s powerful machine learning. Machine learning based on artificial neural networks is currently revolutionising science, engineering and daily life.” . Here you can find the Press Release about the 2024 Nobel Prize of Physics.