Week 1 HW: Final Project Ideas
Week 1 HW: Final Project Ideas
- What you would like to biomanufacture and why?
I hope to biomanufacture an enzyme capable of directly synthesizing the glycyrrhetinic acid derived P-gp inhibitor Sol-DMAP
Between 12.9% and 48% of all schizophrenia patients suffer from a pharmacoresistant form of the disease which does not respond adequately regardless of combinations of any drugs or therapies. The major biological driver of this resistance is P-glycoprotein, an ATP-dependent efflux transporter at the blood-brain barrier that actively pumps a wide range of structurally dissimilar antipsychotic drugs back out of the brain tissue before the drug can act & P-gp function is upregulated in this class of schizophrenia patients. Sol-DMAP is a semisynthetic derivative of the licorice root product glycyrrhetinic acid, which binds to the P-gp active site & inactivates it, restoring drug sensititvity—notably including a 4.7-fold increase in intratumoral accumulation of a co-administered drug in mouse models. While the focus of my interest was in this psychiatric application, as discussed just prior P-gp inhibitors & Sol-DMAP in particular, was investigated for the efficient targeting of tumors during cancer therapies to allow for thorough perfusion of the drug within the tumor while blocking P-gp efflux out by the drug to protect the body from the therapeutic. Sol-DMAP is currently made through a synthetic amide bond coupling reaction
relavent cited work:
- https://doi.org/10.1155/2015/484963
- https://doi.org/10.3390/molecules29204939
- https://doi.org/10.3390/ijms262411912
- What living system would you use?
The chassis here is a TX-TL system for an E. Coli-derived cell-free lysate, where the bacteria are grown, lysed, and the transcription/translation & metabolic machinery are harvested into a reaction mix that can express any DNA template added without needing to transform or culture a new living strain for the many variants we’ll be testing. Also of importance is that cell-free systems tolerate toxic and growth-inhibiting substrates & intermediates such as the plant derived and synthetic chemicals we’ll be working with here.
While I have to do some more research to lock in this starting point in detail, but generally we will be working with enzymes in the ATP-dependent amide bond synthetases (ABS) family of enzymes, whi perform essentially the same reaction we want to perfoem—the adenylation and coupling of a carboxylic acid to an amine in a single step reaction.
relavent cited work:
- How would you use automation?
I believe the best design for this project would be an automated Design-Build-Test-Learn (DBTL) cycle, roughly as follows:
- Design: AlphaFold generates a structural model of the active sites of our original ABS enzyme of interest docked with the glycyrrhetinic acid and the DMAP-amine so that we can identify what substrates would have to be mutated to accomodate the bulky GA molecule rather than via blind random search for mutations and designing.
- Build: the candidate mutations are assembled into a physical mutant gene library. Based on the Python script I’ve prepared, an Opentrons liquid handler loads each mutant along with the GA & amine substrate into its own individual TX-TL reactions (a different variant tested in parallel in each well)
- Test: each reaction in the library expresses its enzyme variant & then goes through a two-stage assay. First, as a proxy for determining if any amide-coupling reaction occured, each reaction is screened through a plate reader to look for ATP consumption / pyrophosphate release / etc. to narrow down the hundreds of variants to a short candidate list, which is then screened via LC-MS to confirm if Sol-DMAP was produced & if so, how efficiently.
- Learn: the activity and mass-spec data are mapped back onto the original enzyme structure to understand which mutated positions helped/hurt accomodation of GA structure. This data can either flow into the next design iteration stage of the loop to refine which residues to target, etc. or it can close the cycle if the first library is successful enough.
- How much would you need to make, and how long would it take?
~100-400 mutant variants per round (depending on 96 or 384 well formats chosen) across approximately two to three DBTL rounds. At production scale (not carried out here, but I think ), it’s hard to say especially since the hypothetical human doses are not really known & would have to be converted up from the mouse models & applied to human doses. However, from our Unit 1 readings, well optimized cell-free reactions run at $90–550 per liter, which we can treat as a baseline for understanding scaling.
For the timeline: a few weeks per Alpha Fold round & a few weeks per whole screen. So ~4-6 weeks per cycle.
- What don’t you know yet?
- I haven’t selected a candidate enzyme from the ABS family yet, & even if I had, I do not yet know if the candidate enzyme active sites can be mutated to accommodate a substrate as large and rigid as glycyrrhetinic acid at all, the proven range is w/ aromatic acids & not triterpenoids.
- Is AlphaFold protein-ligand docking reliable enough to narrow the mutant search, or is a broader semi-random library better.
- Any actually successful variants would probably require optimization beyond the scope of this to actually have reliable conversion efficacy.
- Since this molecule has not yet been produced at scale even if we achieve a working enzyme it’s difficult to estimate whether this biocatalytic route would be prefereable (especially if ramped up to scale) cost-wise once accounting for real yields, purification costs, enzyme production cost, etc.
Important Disclosures:
- The idea for this project comes from a sensibly pared down version of a wildly overambitious (rejected for its scope) proposal for designing an (entirely theoretical) protocol for the designing of candidate enzymes for the structural modification of glycyrrhetinic acid to Sol-DMAP in an entirely in vivo process using E. coli from my previous coursework that I reworked to be both feasible & tailored to the aims and goals of this class. (I must have been so unrealistic w/ that previous assignment i bc i got lulled by not having to do anything & gone a bit pie in the sky, but actually turning the idea into a possible working project focusing on the actually interesting and useful part is a big reason to do HTBAA! 😅 )
- AI usage: to do some basic brain-storming & to explore how possible this idea was I did some very basic prompting of Claude Sonnet 5:
“My original idea to build the whole living cell pipeline in a single semester is obviously not really possible, so the actual project should be to build and automate the testing rig that finds the enzyme maybe? So taking an existing enzyme that is close to doing this job idk what rn & using alphafold to design docking into the glycyrrhetinic acid, then writing an opentrons program & making a bunch of mutant variations of it w/ cell-free txtl, & once we run the assay we hope to find at least one variant successfully makes Sol-DMAP maybe??” (after uploading a .pdf file of my old homework proposal from a previous class)
“summarize the key points of this paper” + the .pdf file for any very long or difficult readings)
- What you would like to biomanufacture and why?
I hope to biomanufacture an enzyme capable of directly synthesizing the glycyrrhetinic acid derived P-gp inhibitor Sol-DMAP
Between 12.9% and 48% of all schizophrenia patients suffer from a pharmacoresistant form of the disease which does not respond adequately regardless of combinations of any drugs or therapies. The major biological driver of this resistance is P-glycoprotein, an ATP-dependent efflux transporter at the blood-brain barrier that actively pumps a wide range of structurally dissimilar antipsychotic drugs back out of the brain tissue before the drug can act & P-gp function is upregulated in this class of schizophrenia patients. Sol-DMAP is a semisynthetic derivative of the licorice root product glycyrrhetinic acid, which binds to the P-gp active site & inactivates it, restoring drug sensititvity—notably including a 4.7-fold increase in intratumoral accumulation of a co-administered drug in mouse models. While the focus of my interest was in this psychiatric application, as discussed just prior P-gp inhibitors & Sol-DMAP in particular, was investigated for the efficient targeting of tumors during cancer therapies to allow for thorough perfusion of the drug within the tumor while blocking P-gp efflux out by the drug to protect the body from the therapeutic. Sol-DMAP is currently made through a synthetic amide bond coupling reaction
relavent cited work:
- https://doi.org/10.1155/2015/484963
- https://doi.org/10.3390/molecules29204939
- https://doi.org/10.3390/ijms262411912
- What living system would you use?
The chassis here is a TX-TL system for an E. Coli-derived cell-free lysate, where the bacteria are grown, lysed, and the transcription/translation & metabolic machinery are harvested into a reaction mix that can express any DNA template added without needing to transform or culture a new living strain for the many variants we’ll be testing. Also of importance is that cell-free systems tolerate toxic and growth-inhibiting substrates & intermediates such as the plant derived and synthetic chemicals we’ll be working with here.
While I have to do some more research to lock in this starting point in detail, but generally we will be working with enzymes in the ATP-dependent amide bond synthetases (ABS) family of enzymes, whi perform essentially the same reaction we want to perfoem—the adenylation and coupling of a carboxylic acid to an amine in a single step reaction.
relavent cited work:
- How would you use automation?
I believe the best design for this project would be an automated Design-Build-Test-Learn (DBTL) cycle, roughly as follows:
- Design: AlphaFold generates a structural model of the active sites of our original ABS enzyme of interest docked with the glycyrrhetinic acid and the DMAP-amine so that we can identify what substrates would have to be mutated to accomodate the bulky GA molecule rather than via blind random search for mutations and designing.
- Build: the candidate mutations are assembled into a physical mutant gene library. Based on the Python script I’ve prepared, an Opentrons liquid handler loads each mutant along with the GA & amine substrate into its own individual TX-TL reactions (a different variant tested in parallel in each well)
- Test: each reaction in the library expresses its enzyme variant & then goes through a two-stage assay. First, as a proxy for determining if any amide-coupling reaction occured, each reaction is screened through a plate reader to look for ATP consumption / pyrophosphate release / etc. to narrow down the hundreds of variants to a short candidate list, which is then screened via LC-MS to confirm if Sol-DMAP was produced & if so, how efficiently.
- Learn: the activity and mass-spec data are mapped back onto the original enzyme structure to understand which mutated positions helped/hurt accomodation of GA structure. This data can either flow into the next design iteration stage of the loop to refine which residues to target, etc. or it can close the cycle if the first library is successful enough.
- How much would you need to make, and how long would it take?
~100-400 mutant variants per round (depending on 96 or 384 well formats chosen) across approximately two to three DBTL rounds. At production scale (not carried out here, but I think ), it’s hard to say especially since the hypothetical human doses are not really known & would have to be converted up from the mouse models & applied to human doses. However, from our Unit 1 readings, well optimized cell-free reactions run at $90–550 per liter, which we can treat as a baseline for understanding scaling.
For the timeline: a few weeks per Alpha Fold round & a few weeks per whole screen. So ~4-6 weeks per cycle.
- What don’t you know yet?
- I haven’t selected a candidate enzyme from the ABS family yet, & even if I had, I do not yet know if the candidate enzyme active sites can be mutated to accommodate a substrate as large and rigid as glycyrrhetinic acid at all, the proven range is w/ aromatic acids & not triterpenoids.
- Is AlphaFold protein-ligand docking reliable enough to narrow the mutant search, or is a broader semi-random library better.
- Any actually successful variants would probably require optimization beyond the scope of this to actually have reliable conversion efficacy.
- Since this molecule has not yet been produced at scale even if we achieve a working enzyme it’s difficult to estimate whether this biocatalytic route would be prefereable (especially if ramped up to scale) cost-wise once accounting for real yields, purification costs, enzyme production cost, etc.
Important Disclosures:
- The idea for this project comes from a sensibly pared down version of a wildly overambitious (rejected for its scope) proposal for designing an (entirely theoretical) protocol for the designing of candidate enzymes for the structural modification of glycyrrhetinic acid to Sol-DMAP in an entirely in vivo process using E. coli from my previous coursework that I reworked to be both feasible & tailored to the aims and goals of this class. (I must have been so unrealistic w/ that previous assignment i bc i got lulled by not having to do anything & gone a bit pie in the sky, but actually turning the idea into a possible working project focusing on the actually interesting and useful part is a big reason to do HTBAA! 😅 )
- AI usage: to do some basic brain-storming & to explore how possible this idea was I did some very basic prompting of Claude Sonnet 5:
“My original idea to build the whole living cell pipeline in a single semester is obviously not really possible, so the actual project should be to build and automate the testing rig that finds the enzyme maybe? So taking an existing enzyme that is close to doing this job idk what rn & using alphafold to design docking into the glycyrrhetinic acid, then writing an opentrons program & making a bunch of mutant variations of it w/ cell-free txtl, & once we run the assay we hope to find at least one variant successfully makes Sol-DMAP maybe??” (after uploading a .pdf file of my old homework proposal from a previous class)
“summarize the key points of this paper” + the .pdf file for any very long or difficult readings