When: July 10-14, 2023
Intended Audience: Graduate Students, Post-Graduate Research and Clinical Fellows, as well as Early-Career Scientists with interest in learning more about Neuroinformatics
Where: In-person at the Krembil Centre for Neuroinformatics, 12th Floor, 250 College Street, Toronto, Canada
Cost: Minimal ($50), only 30 spots available
Click here to apply!
Application Deadline June 2, 2023
Those selected to participate will be contacted in mid-June to register.
Overview
The Krembil Centre for Neuroinformatics is excited to offer a five-day intensive project-based learning week where trainees will learn hands-on techniques for integrating multi-scale neuroscience data. This course is designed to introduce participants to the concepts and methods behind psychiatric neuroinformatics - encompassing genetics, brain structure and function, and cognition.
In addition, participants will uncover the links between modalities of human genomics, neuronal electrophysiology, structural and functional neuroimaging, and observed behaviour that KCNI scientists are integrating through a series of virtual modules and a group-based project using real-world data types to study mental illness.
This unique learning opportunity will prepare participants to handle and analyze multiple data types in hopes that their own research may benefit from collaborative, multi-modal approaches. Critically, participants will also learn about best practices for data management and quality control in the context of integrative analysis.
Registration now open Click here to apply!
Application Deadline June 2, 2023
Those selected to participate will be contacted in mid-June to register.
Course Requirements and Prerequisites
Applications from senior undergraduate students, graduate students, post-graduate research and clinical fellows, and early-career scientists will be considered. Researchers from diverse backgrounds (e.g. medicine, computer science, biology, psychology, engineering etc.) are encouraged to apply. To ensure that all attendees can fully follow and benefit from the practical assignments, fundamental and demonstrable experience in R and Python is a minimum requirement.
Delivery Method
Students will be provided with a collection of virtual didactic teaching (lectures) and hands-on tutorial components to engage critical thinking and develop practical skills in crucial selected areas. Lessons will be led by members and affiliates of the KCNI team, including faculty at the University of Toronto’s Department of Psychiatry. Successful applications to Project Week will be mentored (in-person or hybrid) by KCNI Scientists as they complete one-week intensive group projects matched to their interests. These projects will allow participants to build mastery in selected integrative research methods.
All project week participants will take part in each full day of training. Students will have the opportunity for project-specific discussions, collaboration, and guidance outside of structured time.
Students will present their group project to KCNI Scientists at the conclusion of Project Week. Project week will occur in person at CAMH KCNI and be supplemented by multiple networking, collaboration, and socializing opportunities.
This year's projects
This years students will be split into groups who will each focus on one of the projects given below, will guidance from KCNI trainees and staff
Lead Scientist/Lab: Shreejoy Tripathy
Teaching AssisstantsAssistants: Mel Davie & Dan Kiss
Project 1 - Single Cell Transcriptomics
Integrative analysis of mouse and human single-cell gene expression data. What are conserved cell types in the brain and what are their characteristics?
Main idea: perform integrative analysis of mouse and human (and possibly other species, including macaques and marmosets) neocortical cell types according to transcriptomics and possibly intrinsic electrophysiology.
Key questions:
- Can we identify orthologous cell types between species?
- What genes / features distinguish cell types?
- Are there aspects of these features that seem relevant to computational processing of cells and circuits?
- What characteristics of these features can be modelled in the context of cell and circuit models?
What (dataset) resources are available to help answer this question? Allen Institute for Brain Sciences Cell Types database https://celltypes.brain-map.org/
Programming Languages: R/Studio, (Seurat)
Link to GitHub: https://github.com/sonnyc247/KCNISS_2022_Week2
Lead Scientists/Labs: Etay Hay and Andreea Diaconescu
Teaching AssisstantsAssistants:
- Frank Mazza,
- Alexandre Guet-McCreight,
- Colleen Charlton,
- Milad Soltanzadeh
- Zheng Wang
Project 2 - Stratifying risk for schizophrenia using in-silico cell-specific EEG biomarkers
The project will apply novel EEG biomarkers of cell-specific cortical inhibition (from parvalbumin interneurons) and underlying cognitive precision/prediction errors, to stratify schizophrenia risk and severity.
The biomarkers are computed from features of EEG signals measured during auditory oddball (deviant vs standard stimulus) processing, and will be also linked to resting-state EEG features such as the power in different frequency bands and aperiodic/periodic components of the EEG. The project will involve manual analysis as well as machine learning (deep learning) classification.
Learning outcomes:
- Analysis of task and resting-state EEG
- Traditional and new EEG analysis methods
- Cell-specific signatures of EEG from the cortical neuronal network
- Computational modeling of brain processing underlying the EEG signals using hierarchical Bayesian models (e.g., the hierarchical Gaussian filter)
- Trial-by-trial modelling of EEG data using statistical parametric mapping and dynamic causal modelling
- Machine learning (deep learning) classification of positive symptom severity
Significance:
These types of analysis seek to improve schizophrenia patient stratification and early detection in at-risk/young population.
Programming languages: Python, Jupiter notebook, MATLAB (MATLAB access will be provided for all course participants free of charge)
Reference reading:
Mazza et al 2023, PLoS Computational Biology
https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1010986
Charlton et al 2021, Schizophrenia
https://www.nature.com/articles/s41537-022-00302-3
Lead Scientist: Daniel Felsky
Teaching AssisstantsAssistants:
- Earvin Tio
- Rajith Wickramatunga
- Mohamed Abdelhack
- Emily Wiljer
- Mu Yang
Project 3: Transdisciplinary Mental Health Modeling with Machine Learning
From domain-specific data to transdisciplinary models: an exercise in whole-person modeling from feature engineering to model explanations.
What's this project about?
Building a transdisciplinary machine learning model to predict risk for mental illness in a large simulated human population dataset. This will involve exposure to genetic, neuroimaging, and sociodemographic data types and their integration.
Key Questions/Objectives
- Objective 1: How to approach a multi-disciplinary analysis
- Steps to approaching a new dataset: how to handle data formats, variable types, missing data, transformations?
- Objective 2: Feature engineering from different data modalities
- How do we go from raw data (i.e. genotype, brain scans) to individual level features?
- Objective 3: develop a transdisciplinary machine learning model
- How do we combine multiple data types into one predictive model?
- Can we analyze feature importance to understand relative contributions from different domains?
Programming Languages
- Bash, Python, R
Link to code: TBA
Lead Scientist: John Griffiths
Teaching Assistants
- Davide Momi
- Shreyas Harita
- Zheng Wang
- Andrew Clappison
- Sorenza Bastiaens
- Taha Morshedzadeh
- Kevin Kadak
- Parsa Oveisi
Project 4 - Whole-brain modelling of noninvasive neurostimulation therapeutics and neuroimaging connectomics
What’s this project about?
Main idea: use neuroimaging-based computational models of TMS brain stimulation to investigate the micro-physiological and macro-network effects of TMS therapies used in psychiatry and neurology
Key questions:
- What do the electromagnetic stimulation patterns look like in 3D volume space and 2D surface space for different TMS target regions and targeting methodologies (eg 10-20 coords vs beam-F3 vs neuronavigated)?
- What do the associated whole-brain network stimulation patterns look like, as calculated using anatomical and/or functional connectivity information from MRI scans?
- What is the spatiotemporal pattern of stimulated activity propagation across the brain as measured with source-localized EEG?
- How do we use connectome-based neural mass modelling to unpack the structure of this activity propagation?
- How do neurophysiological properties such as excitability and level of inhibition affect this propagation, and differ across individuals and following clinical interventions?
- How can we incorporate theories and models of neural plasticity into the picture here?
Programming Languages: Python, PyTorch
Link to GitHub:
The project will be based on the code and associated (open-access) scientific papers available at these two github repositories
https://github.com/GriffithsLab/PyTepFit
https://github.com/GriffithsLab/HaritaEtAl2022_tms-efield-fc
Reference reading:
Momi et al. ELife 2023
Harita et al. Front Psych 2022
Griffiths et al. 2022
Griffiths et al. 2020