Epigenetic (Bayesian?) Gene Training

[Cathal Garvey]

Hi all,

So, in the Cold-Resistant-Wheat thread, I suggested that costs in plant
projects may limit ones' ability to optimise expression patterns in a
plant for different climates, and that this can impose a yield burden or
make the trait less valuable. This kind of trade-off sucks; at best, it
leads to many varieties of the same trait, each optimised for different
conditions, and at worst it means you're only optimised for a single
climate.

So, in parallel to all this I'm playing around in machine learning, and
there's a host of different algorithms that are designed to "train"
under different conditions and then operate in "fixed" mode to classify
things or generate a binary output.

So, I'm wondering; structurally, how might one engineer an epigenetic
system into ones' synbio projects so that 'training' under lab
conditions can result in a pretty stable phenotype?

A blunt instrument method would be to have many copies of an expression
regulator or sigma factor or repressor, each with different
methylation-sensitive promoters, and to arrange for these promoters to
be methylated or demethylated to tweak the amounts of
regulator/sigma/repressor generated as a base-line. More dynamic
regulation patterns could be used as an afterthought in each and every
case, so you can imprint the basic promoter strength and *then* apply
regulation cascades as usual.

I'm wondering if we could do something more elegant, though; anyone
looked into this?
--
Scientific Director, IndieBio EU Programme
Now running in Cork, Ireland May->July
Learn more at indie.bio and follow along!
Twitter: @onetruecathal
Phone: +353876363185
miniLock: JjmYYngs7akLZUjkvFkuYdsZ3PyPHSZRBKNm6qTYKZfAM
peerio.com: cathalgarvey

[theiman]

Hi Cathal,

If there is any gene expression data available, you may want to look at the following paper: a statistical approach to virtual cellular experiments- improved causal discovery using accumulation IDA by Franciscan Tarittis, Rainer Spang, and Julia C Engelmann in bioinformatics..

Sincerely,

Tom