January 2018

Week 1: 1/8-1/12

Apprentice Maggie and Brandon
- Carve out some time to work with Brandon
Abstracts for the symposium due Feb. 9th
Outside Conferences
- One at St. Mary's in the East Bay
- Tri Beta Pacific District Convention - if SoCal, then probably this one. Saturday March 17th?
Priority levels: 0 > 0.5 > 1 > 2
Will be working on deleting db5 network edge by edge; 28 runs
- LSE/minLSE to see what the effect of the deletions are
- run models, get gephi stats
Brandon show me how to do post-hoc analysis
on team meeting issue, post schedule of when I will be working (4 hours)

running db5 with missing data and no missing data
- start 1/16, figure out if there is some type of variation/different network weights between missing values and no missing values
- if it works, make workbooks by next Tuesday.
db5 deletions
- worried about deleting:
  - ZAP1-ACE2: delete ZAP1 (when you delete the edge you also have to delete the node.
  - GCR2-MSN2: delete GCR2
  - YHP1-GLN3: delete GLN3
- db5 with and without missing values: graph looks the same in GRNsight
edge by edge
- today
  - missing/non-missing
- future
  - 28 model runs on GRNmap (next week)
  - 28 gephi stats (next 2 weeks)
  - LSE: min LSE ratios GRNmap output (1-3 weeks from now)
  - compile graph stats into a single excel file
    - compare each of the edge deletions and see if the overall network is statistically significant (sig. difference between original network and network with deletion, if there is a significant difference then the edge that was deleting is theoretically important)
started 3 runs
- missing data, no missing data, and missing data on beta (no edge deletions, just db5)
met with Brandon on Wednesday, went over general things about what has happened in the project.

Does a spreadsheet formula for LSE/minLSE exist?
Maggie and I ran the 28 iterations of GRNmap to conduct an edge deletion analysis, also ran generated output workbooks through Gephi
Metagraph?
Delete 2 edges to see if the LSE/minLSE ratio decreases more?
- When we delete things, we delete uncertainty. Must make sure we have a good control
- run a series of random networks with the same number of nodes and edges to make a comparison
Abstract deadline for LMU Symposium is the 9th
Using other databases from Biomathematical Modeling/other class?
Create a network that doesn't have any of the current genes in it?
For next week: finish up what we have been working on; not running any new models

can't use a paired t-test to do the analysis - what type of ranked test can we do?
- keep getting a divide by 0 error because in some cases the data are equal
  - transcription error, or deleting that edge did nothing