Decoding Higher-Order Cognition from Invasive Neural Signals

MaCSBio Lecture Series

The decoding of higher order cognition directly from recordings of neural activity in the brain could enable a new generation of prosthetic devices. Accurate information about memory processes, reward perception and attempted speech and motor activity will allow targeted interventions and next-level human-computer interaction. In this presentation, I will present work with neurological patients that have electrodes implanted deep into their brains for clinical procedures. By piggybacking on these clinical routines, we are able to record high-fidelity neural activity across a variety of brain areas and align them to cognitive tasks. Through the application of machine learning, we are able to decode higher-order cognition from these recordings and process the output in real-time.

However, current technical limitations for deconvoluting complex protein mixtures and thorough mapping to disease targets have prevented their widespread utilization as a therapeutic treatment. Juvena Therapeutics is overcoming these technical challenges by building a drug development platform that enables systematic investigation of the therapeutic potential of secreted proteins by integrating quantitative proteomics, transcriptomics and machine learning. Candidate proteins are selected from a proprietary disease-modifying protein library derived from human embryonic stem cells using mass spectrometry and transcriptomics. Candidate rank ordering is achieved through machine learning models and bioinformatics leveraged from biological, biophysical, and experimental features. In vitro phenotypic screening and preclinical validation of candidates is accelerated with using deep learning models predicting the cell state, tissue regeneration, and the cognitive behavior of rodents. This platform has identified multiple lead hits across different therapeutic indications. Our goal is to develop tools that enable identification of disease modifying secretomes from other cell types as well as annotating isoform variants and post-translational modifications. Ultimately, Juvena Therapeutics is creating a knowledge graph of regenerative protein biology in which protein ligand-receptor interactions are mapped to a phenotypic response for a given diseased tissue.

About the speaker

Christian Herff is an Assistant Professor in the Department of Neurosurgery at Maastricht University where he heads the Neural Interfacing Lab. His research interests lay in the decoding of higher-order cognition from neural signals to combine natural with artificial intelligence. Christian has a background in Computer Science, which he studied at the Karlsruhe Institute of Technology, IIT Delhi and NTU Singapore. He obtained his PhD in computer science from the University of Bremen. His research attracted national and international funding including an NOW VENI in 2019.