HC11 Genetics Flashcards
Imaging genetics
- Combines brain imaging and genetics
- Identifies genetic variants (genotype) linked to differences in brain structure & function (phenotype)
- Examines how genetics relate to behavioral traits (e.g., cognition, psychiatric disorders)
DNA
Deoxyribonucleic Acid
Spiral-shaped, complex molecule with base pairs
Contains genetic instructions
Gene
- Segment of DNA that codes for a specific protein
- Located on chromosomes
Chromosome
- Long strand of coiled DNA
- Contains multiple genes
Polymorphism
Variation in a gene or DNA segment
Single Nucleotide Polymorphism (SNP)
- Variation of a single base pair in the genome
- Some individuals may have G, others A at a specific location
- Can result from mutations, cell division, or sexual reproduction
Family linkage studies
- Identify genes in families with high disease prevalence
- Success: Diseases caused by a single gene mutation (e.g., cystic fibrosis)
- Failure: Common diseases with multiple interacting genes
Twin studies
Compare monozygotic (MZ) twins (100% genetic similarity) and dizygotic (DZ) twins (50% genetic similarity)
Discordant MZ Twin Design
Discordant means that the trait is not shared between the twins.
For example, only one twin scores high on ADHD, OCD, generalized anxiety, major depression, etc. These differences also reflect environmental effects.
Candidate-gene approach
- Tests specific hypotheses linking gene variation to brain structure/function
- Focuses on one or more SNPs
Strength: High statistical power
Weakness: Cannot discover new genes or interactions
Genome-wide association studies (GWAS)
- Examines large sets of SNPs across the genome
- Can identify unknown genetic links
Weakness: Low power due to multiple comparisons
Consortia & Meta-Analytic Approaches
- Multi-center consortia: Large datasets from multiple centers (e.g., IMAGEMEND)
- Meta-analyses: Combine findings from different studies (e.g., ENIGMA)
Polygenic Risk Scoring
Polygenic risk scoring looks at many small genetic differences (SNPs) across a person’s entire DNA. It combines them into a single score using data from previous large studies (GWAS). This score shows how a person’s genetic risk for a certain disease compares to others. The score is based only on genetics and does not depend on other factors you are studying.
Polygenetic risk scoring - advantages
Maximization of power even with small samples. - Replicate findings.
Computational modeling
Uses math, physics, and computer science to study behaviour of complex systems. It wants to find mechanistic explanations of how the nervous system processes information to give rise to cognitive function and behaviour using computer stimulations
Why use computational modeling?
- Test theories of cognition
- Integrate knowledge from multiple fields
- Observe complex interactions
- Generate empirical predictions
- Perform artificial lesioning to test validity
Computer simulations
An imitation or reproduction of behaviour in an alternative medium, simulated cognitive processes. These simulated cognitive processes form the model is called artificial intelligence.
Computer simulation - simplify
Simulations allow you to simplify things to make it easier to look at how things work. If you wouldn’t do this, then it is very hard to examine the human brain. The human brain by itself is too complex to simulate. An equally complex model not helpful: not easier than studying the brain.
Computer simulations - animate
Dynamic presentations: they are able to capture and reveal how dynamic processes unfold over time.
Computer simulations - emerge
study how psychological properties emerge from brain-like models.
Why do we build models?
- explicit hypotheses and assumptions necessary to test theories of cognition.
- provides a framework for integrating knowledge from various fields, you can bring all the knowledge together within one model.
- possible to observe complex interactions among hypotheses.
- provides ultimate control.
- leads to empirical predictions.
- artificial lesioning makes it possible to test a model’s validity, if you make a lesion in the model, you can see the effect.
Neural network
A biologically inspired computational model patterned after network of neurons present in the human brain. It has different levels of complexity, from single neurons and synapses up to abstract connectionist-type or population-level descriptions of neural networks.
The basic form of a neural network has 3 different layers:
Input Layer: Receives information
Hidden Layers: Processes information
Output Layer: Produces results
Deep neural networks (DNNs)
A neural network with multiple layers between the input and the output layers, so a network with more hidden layers.
Constraints to make the neural network biological plausible:
- Multi-layered neural architecture.
- Each processing node (which reflects a neuron) is connected to many others.
- Each connection (which reflects a synapse) is characterized by a connection weight.
o positive values = reflect the degrees of excitation.
o negative values = reflect the degrees of inhibition. - Each simulated neuron receives multiple inputs (which reflect dendrites).
- If the sum of the inputs is above the threshold: an output is triggered from the receiving node (according to a transfer function).
- Memory and learning depend heavily on changing connection weights (which reflects synaptic properties) → experience-dependent plasticity mechanisms.
Supervised learning
Develop a predictive model based on both input and desired output data. Input and output is given so known. The weights are adjusted until the output has the desired value: classification/regression. The idea is that training data can be generalized, the model can be used on new data. All information is given, input and output.
Unsupervised learning
Group the data and interpret the data based only on input data. A model that might have generated that set: clustering. You get data (input) without any information (labels) and you don’t know the output, the output becomes a structured dataset by using clustering.
Reinforcement learning
the value of the output is unknown beforehand but the network provides feedback whether the output is right or wrong: sequential decision problem / trial-and-error. This is done to train animals.
Human brain project
Using simulation of multi-scale models to uncover the neural mechanisms underlying cognitive processes, such as learning, perception, sleep and consciousness. Construct neural network simulators from neuromorphic chips that use transistors to simulate neurons and synapses.
Advantages of neuromorphic computing
- Energy efficiency.
- Execution speed.
- Robustness against local failures.
- Ability to learn, it is possible to access the different timescales involved in learning and development.