Here's a link to the preprint if you made it this far:
www.biorxiv.org/content/10.1...

any feedback would be highly appreciated!

A big thank you to all co-authors - looking forward to seeing where we can take this approach.

If you are interested in joining us on this effort, check out our website: www.pinglay-lab.com

Finally, we demonstrate that data resulting from SGE is compatible with training predictive machine learning models.

We are very excited about using SGE to generate the synthetic data needed to train the next generation of models for biological design.

SGE is generalizable across cell types. We engineer T-cells (Jurkat) to grow without valine.

We believe a similar strategy could help create more resilient T-cells for therapy, capable of surviving and functioning in the metabolically depleted environments of tumors. Hopefully more here soon!

Here's a link to the preprint if you made it this far:
doi.org/10.1101/2025...

any feedback would be highly appreciated!

A big thank you to all co-authors - looking forward to seeing where we can take this approach.

If you are interested in joining us on this effort, check out our website: www.pinglay-lab.com

Finally, we demonstrate that data resulting from SGE is compatible with training predictive machine learning models.

We are very excited about using SGE to generate the synthetic data needed to train the next generation of models for biological design.

We then used SGE to engineer CHO cells to grow without isoleucine, a feat we could not achieve via rational design and delivery of entire synthetic pathways.

Again, mitochondrial localization was favored, with individual clones reflecting ~40-50kb of integrated DNA!

Using SGE, we screened millions of pathway combinations in a single experiment to engineer CHO cells that grew at WT rate (~1.1 day/doubling) in valine-free medium.

Intriguingly, the best clones all employed mitochondrial localization of pathway components, not cytoplasm as in our prev. design.

In collaboration with Harris Wang’s lab, we previously engineered cells to grow without valine by importing 4 genes from E.coli.

However, the cells grew 4x slower than normal - and we could not extend this strategy to enable any other amino acid prototrophies.

doi.org/10.7554/eLif...

As a test case, we used SGE to engineer essential amino acid prototrophy in mammalian cells, a behavior last seen over 500 million years ago.

Unlike E. coli, which can make all 20 proetinogenic amino acids, mammals lack the pathways for 9 “essential” ones and must obtain them through the diet.

In SGE, we clone and deliver a TU library at high MOI so that each cell gets a random mix, assembling a unique synthetic metabolic pathway per cell. Cells with the desired phenotype (e.g., survival or fluorescence) are selected, and TU barcodes are sequenced to identify functional combinations.

To address this, we developed Shotgun Genetic Engineering (SGE), which leverages the fact that building and delivering many small, barcoded transcription units - each with a gene, promoter and localization signal - is exponentially easier than delivering a single large construct to a mammalian cell.

Mammalian metabolic engineering is key to advancing bioproduction, cell therapy, and rejuvenation.

But as pathway complexity grows, so does the combinatorial design space! However, delivering large DNA constructs to mammalian cells is inefficient, making large unbiased screens intractable.

Such a cool story Maximus! Congrats

Let’s gooo! Congrats Dave and team - glad to see this out.