1. Reachability Analysis
Markus Krummenacker
Bioinformatics Research Group
SRI, International
August 25th, 2009

2. Inputs & Outputs Growth Medium: The set of Starting Compounds
Biomass Composition: The set of Target Compounds
Auxiliary Nutrients: 2 kinds.
Dead ends (dead starts)
Bootstrapping circular pathways
Outputs: Set of reactions that were reached (“fired”), Set of compounds produced

3. Basic Algorithm
Given the starting compounds and auxiliaries, “fire” all reactions that have all inputs satisfied.
Add all newly produced compounds to the set of compounds that are available.
”Fire” next set of reactions, and repeat, until no further reactions can be “fired”.

4. GUI for Reachability “Experiments”
Tools->Reachability Analysis...
Manages time-stamped files of Reachability runs.
The last file is actively editable.
The prior files are a record of what was tried and with what result. Such files can be duplicated.
Growth Media and Biomass Compositions are frames in the PGDB.
The Auxiliary Nutrients are stored in the file.

5. Running a Reachability “Experiment”
Button: Invoke Reachability Analysis
To visualize results, button: Display on Cellular Overview
“Fired” reactions are highlighted in yellow.
Mouse-over on Reactions show “fired” status, and which compounds are available or absent.
This allows investigating gaps in the network.

6. Some Limitations
Auxiliary Nutrients, needed for Bootstrapping Pathways that need a compound, before they can produce more of it. Common ones are: ATP, NAD+, CoA
Reactions formulated in terms of compound classes
Polymerization Pathways

7. Class Expansion to Instances
Reactions formulated in terms of compound classes.
Example: |Xs| + H2O = |Ys|
|Xs| is a class with instances X1 X2 X3
|Ys| is a class with instances Y1 Y2
Expansion code tries to pair all instances on LEFT and RIGHT sides with each other, substituting for the class
Test whether for a given instance in |Xs| , there is only 1 instance in |Ys| that leads to a mass-balanced reaction equation. If yes, create the instance-based reaction on the fly.
(No chemical structure matching yet.)

8. Other Class-Instance Mappings
To bridge gaps that can’t be bridged automatically, some hard-coded mappings exist.
Half a dozen mappings set up by (initialize-pseudo- reaction-mappings)

9. Polymerization Pathways
Metabolites undergoing polymerization are represented as classes (the structure having an R group).
The hallmark of a polymerization pathway is that 1 reaction, the polymerization step, is unbalanced.
As a hack, the chemical formula of the misbalance is determined, and represented as a pseudo-compound, on the fly.
This pseudo-compound can then be used for balancing some of the hard-coded Class-Instance mappings, by adding an integral multiple of the pseudo-compound to the mapping reaction.