W6L2

Question 1

Why we make gene network

Answer 1

• Networks are more representative of reality
• Some types of networks:
• Gene regulatory networks
• Protein-protein interaction networks
• Gene co-expression networks
• Continued shift from understanding the function of a single candidate gene to evaluating cellular activity holistically

Question 2

terminology for gene networks

Answer 2

• Node: A point in the network, in this case a gene or protein.
• Edge: A connection between nodes, an association or interaction

Question 3

Gene regulatory networks (GRNs)

Answer 3

• GRNs capture the regulatory relationships between sets of genes (causal relationship)
• Building a GRN requires knowledge of underlying biology so we can assign directionality

Question 4

Where does GRN take info from

Answer 4

• Hybrid screens
• More general perturbation assays
• Knock-out screens
• Time-course data

Question 5

Core element of GRN

Answer 5

-positive feed back loop, negative feedback loop, flip-flop, feed forward loop
-Each of the core network have different property and outcome
-Understanding core regulatory networks lets us model events inside a cell, make predictions about a future state, and design circuits of our own.

Question 6

Multiple inputs allow for simple logical operations

Answer 6

Combinations of binding factors (e.g.: multiple enhancers… hey…) allow for robust regulatory circuitry
* The simplest models treat this as a Boolean process, but approach can be extended

Question 7

What does negative control of many gene do

Answer 7

-odd number will oscillate while even number one would have a stable expression peek

Question 8

protein-protein network

Answer 8

• Capture direct interactions between proteins
• Generated from experimental data that measures physical interactions between nodes
• Yeast 2 hybrid screens are one of the best sources of information for this
• Unlike GRNs, make no assessment on direction of effects

Question 9

Building protein protein with a lack of biological info

Answer 9

• Gene co-expression networks can be built from genome-wide measurements of gene expresssion across many individuals/conditions
• Much like QTL mapping, require limited understanding of underlying biological process

Question 10

Gene co-expression network

Answer 10

• Edges no longer imply a direct relationship!
• Show only edges above a certain correlation threshold, ignore all interactions below thresholds
• Weight edges more if correlation is higher.
• Ignore direction of the effect (negative vs positive); an unsigned network

Question 11

How to work with messy network

Answer 11

Decompose them into modules (subnetwork):
* Subsets of genes that all behave similarly
* Module identification also doesn’t rely on knowledge of the underlying biology
* Analysing networks suggests a few ways in which to incorporate biology in our insights

Question 12

Comparing networks across groups or time

Answer 12

• Can test for differences in network/module behaviour between states/groups/time points
• Similar intuition to testing differences in gene expression/protein abundance/rate of transcription/etc
• Many of these comparisons are ‘hypothesis generating’ as opposed to ‘hypothesis testing

Question 13

Interpreting network, interpreting traits

Answer 13

• Guilt-by-association approaches let us make guesses about the function/role of a gene on the basis of the rest of the module
• use eQTLs to prioritize putatively casual gene
  -combining eQTLs and undirected co-expression interaction networks for interpretation of GWAS

Question 14

Advantage and disadvantages of Modeling biology with network

Answer 14

• Network approaches enable the description of cellular and genetic processes at great depth
• We don’t need to know much biology to describe some biological processes
• But interpretation is far more challenging!