Click to read the piece on substack (this is also how I share summaries of my reseach - and will post many more things in the future) open.substack.com/pub/ksechkar...

So how can we reframe 'balancing innovation with safety'? Well, it's still vital to *innovate safely* and have robust regulation, but I'd say we also have the responsibility to *innovate FOR safety* - and this premise is baked into the idea of Engineering Biology itself 6/6

Identifying these behaviours, we can then design biosystems with robustness, safety and reliability in mind (e.g. that's what our ongoing @eebio.bsky.social
programme is all about). Innovative stuff, but hardly pitted against safety! 5/6

Using automatic culturing platforms, we work to characterise biotechnologies' behaviours as they're affected by factors like randomness, evolution or (my own area of interest) the sheer interconnectedness and complexity of living systems - so they won't catch us by surprise 4/6

I, for one, am doing a PhD supervised by Prof Harrison Steel at the Oxford Control Group. Just like EngBio's name would suggest, we approach problems in *biology* with principles from control *engineering* - which include precisely robustness, predictability and safety! 3/6

One common point in these discussions is the need to 'balance innovation with safety'. Fair! As mentioned in the debate, one mishap can tarnish the field's reputation for good, so robust regulation is vital. But why do we often frame innovation and safety as opposites? 2/6

Thanks to my supervisor Prof Harrison Steel , as well as to our collaborators from @eebio.bsky.social - Sara Brancato, Prof Lucia Marucci and Dr Ludovic Renson for advice. They’re also working on exciting applications of CBC right now, so watch them closely (what a time to be alive!) 9/9

For now, this is more of a pitch with an in silico proof-of-concept, but we discuss which developments in SynBio and control will help us apply this protocol in vivo. Briefly, advanced control algorithms + single-cell culturing + optogenetics = fun 8/

We successfully simulate our protocol using different models of various complexities (e.g. see results for a mechanistic resource-aware cell model in the pic below) – gives us reason to think it should work in (even more complex) real cells! 7/

From start to finish, our proposed method involves first establishing the probe’s properties, then using it to do CBC with the genetic module of interest. The results can then predict how any two characterised modules will compete for resources! 6/

Here, the module of interest’s output is its fluorescent gene expression, and the feedback input is e.g. chemical induction of another ‘probe’ genetic module competing with it for resources 5/

We suggest using control-based continuation (CBC), where a stabilising external feedback input drives a system through the entire range of its equilibria! 4/

Yet arugably, multistability & bifurcations are the most fun bit! And we know resource competition may cause them (and sometimes that’s even leveraged by circuit designs like our recent Punisher doi.org/10.1098/rsif... ). So characterisation of competition should include them… What should we do? 3/

To gauge resource demands, we often look at how modules compete for gene expression resources with constitutive fluorescent reporter genes. But there’s a problem: in multistable systems, constant resource competition may drive your system to just one steady state among many possible

Hope I’ve gotten you interested enough to go and read our paper for details! And the best way I can conclude is by thanking the editors & reviewers, @Harrison_Steel for supervision and coauthorship, and @Andreas_Porse and other Steel Lab members for their great advice! 10/10

We also consider beneficial *native* gene mutations. For the same growth advantage, these are less prone to triggering the Punisher than synthetic gene loss, yielding better fitness. Clonal interference can thus help to hinder engineered cell populations’ function loss 9/10

How about transient reductions in burden due to uneven plasmid and cell cycle fluctuations? We can likewise model these to show that the Punisher is robust to such disturbances 8/10

Though the Punisher’s aimed at penalising mutations of genes burdening the cell primarily via resource competition, we also model metabolic burden. Turns out, our design can punish mutations of metabolically burdensome genes, too – just tune the chemical induction again! 7/10