Minutes 20101018 BioCompiler

From SnOwy - Ed's Wiki Notebook

Minutes for the Monday morning BioCompiler Meeting; I've only started taking notes here after Brandon's main discussion item

Part of the BioCompiler series of notes --

• Previous: Storm 20101017 BioCompiler -- brainstorm in preparing for this meeting.
• No meeting for 20101011 -- thanksgiving Monday.
• Previous: Minutes 20101004 BioCompiler -- one meeting ago.
• Previous: Notes 20100927 BioCompiler -- two meetings ago.

Contents 1 Search Algorithm 1.1 Andre's Musings about Complexity 2 Project Auditing 2009 3 Black Board 3.1 Agenda 3.2 Dijkstra's Algorithm 4 Fiona's Notes 5 Homework

Search Algorithm

• we can use something like Dijkstra's algorithm to bound our search
• because we are essentially naive about the search algorithm to use, we use a master algorithm to select the best search at each iteration

Andre's Musings about Complexity

• we may be able to estimate the number of parameters required for a given circuit if we know the complexity of the target output
  • Estimate: guess at
  • Parameters: physical circuit components
  • Complexity: number of states required for correct operation
  • Target output: a behaviour of the system
• Ed: I suspect that this actually becomes asymptotic -- hyperbolic with respect to the complexity of the target output
• in the first Dijkstra's Algorithm black board below, there's a square-wave sketch at the bottom of the image
  • this was Andre relating the number of states of output to the number of anticipated components
  • we for instance, shouldn't anticipate more than two components to create that wave -- one activator, one inhibitor

Project Auditing 2009

As seen in Workflow 20101002 BioCompiler

• Fiona: https://docs.google.com/document/edit?id=1QlHgSEmp2y4MKGqxoGgNYk-3KNUHeYUO_aX4KsgT3fA&hl=en&authkey=CIij1d4E
• Matthew: https://docs.google.com/document/edit?id=1_dLR75d7DlWmru2O1jG0msRQRltN2SpPiEHNgo5LlRU&hl=en&authkey=CKaah_II
• Eddie: Notes 20101021 BioCompiler Homework

Black Board

Agenda

• Agenda
  1. Discuss other teams' work (the audits we did)
  2. Discuss abstraction
     • What levels of processing and compilation do we need to consider?
  3. Biocompiler inputs and outputs
• Inputs and Outputs ...
  • A: Biocompiler takes descriptions to circuits
  • B: Circuits takes reactants to products
  • C: Biocompiler takes circuits to protocols
• From Fiona and Andre's audits -- identified the following patterns
  1. Biochemical pathways
     • Boolean Off
     • Tagging
     • Wire OR, Wire AND
  2. Gene Conversion
  • Inputs may include light sensors
    • Example: A stress is ultraviolet light which activates physiological responses.
    • Such inputs -- physiological responses -- are intrinsic to the cell and report a cellular state.

Dijkstra's Algorithm

Brendan describes an idea of how to search the solution space

• a master search algorithm is used which nests several worker search algorithms in parallel
• each worker search algorithm combs the search space differently
• the master search algorithm allows each worker to perform one iteration of search at a time
• the workers must report a fitness score in anticipation of their performance in the upcoming subspace
• the master then allows more iterations for those workers that report greater fitness

• a worker function function must enclose a fitness function that conforms to the above prototype
  • these are the meanings of the above parameters as I remember them -- please correct me if I'm wrong.
  • input: the subspace corresponding to one iteration of the worker search algorithm (?)
  • output: a reference to the output object (?)
  • tree: the remaining subspace to search (?)
  • complexity: the evaluated space and time complexity that the function estimates for the tree above (?)
  • runtime: an estimated physical time required to traverse the tree above (?)
  • memory: an estimated physical memory required to traverse the tree above (?)

Fiona's Notes

Fiona's generously donate her notes in lieu of my own.

iGEM Meeting Oct 18th 2010

Outline:
0. Other Teams
1. Abstraction
1a. Biocompiler: description-> circuit
1b. Circuit: reactant-> product
1c. Biocompiler: circuit-> protocol

i.Biochemical pathway
boolean off
wire OR, AND
tagging (eg. GFP)

ii.Software make up
vector of inputs, receive a vector of outputs
look at the most often used genes first
assume humans have done a good job of picking blocks in the past-> Djiktrasa's algorithm
master- search algorithms, best results propogate

iii. Definitions
synthetic biology vs. biochemical biology-> we have to remember which groups we are writing for

2. Biocompiler
Inputs & Outputs

Next:
define a search space, what are we searching through?

Homework

• Brandon will continue to work on the algorithm logic
• Eddie will continue to interpret what candidate inputs are
  • especially curious are behaviours like "oscillation"
• please feel free to continue auditing the 2009 projects and to brainstorm and learn
• next meeting at the same time on Monday