iiinova — where your digital self keeps evolving

To push further, we analyzed three previously published prospective docking campaigns (AmpC, D4, σ2) from the Shoichet lab. Hundreds to >1,000 molecules synthesized and tested. Co-folding does not consistently improve separation vs docking alone.

140

Test 1: 557 ligand-protein complexes. More often than not, co-folding gets the pose right (AF3 > Chai > Boltz). Docking doesn’t do quite as well (~40%). This is judged by the 2Å standard, which I think is pretty generous. AF3 has a sizable sub-Å population, which is amazing!

132

There is also a feeling that the information within the various layers of the YAAFC architecture should reflect not only structure, but also affinity. This idea is reflected in the “affinity module” of Boltz-2, and likely other models to come...

121

James Fraser @fraserlab.com · 12/09/2025

A lot of the recent excitement around methods like AF3 (and followup methods which I often joking call YAAFCs: Yet Another Alpha Fold Clone) is the promise of modeling ligand-bound complexes from only a protein sequence + a small molecule SMILES string (movie from Charm Therapeutics)

132

it's not my style - but it does have a place for those in less protected positions. Luckily @prereview.bsky.social has a great mechanism for this! See the persistent pseudonyme option:

120

James Fraser @fraserlab.com · 09/09/2025

Can’t square that with:

100

James Fraser @fraserlab.com · 08/09/2025

My view: posting a preprint is already an explicit invitation for community feedback. Authors shouldn’t have to opt in - especially to hear when ORCID-authenticated reviews are posted on some of their trusted partners that they include on their reviews and context side bar

2140

These results suggest that stabilizing Orf9b’s (lipid-bound) dimeric state could be a future therapeutic target. Disrupting lipid interactions might bias Orf9b towards a form that can’t bind Tom70, restoring immune activation.

Proposed model of the Orf9b-Tom70 equilibrium
Orf9b transcripts are translated to produce monomeric Orf9b proteins that can bind to Tom70
leading to a suppression of IFN. Monomeric Orf9b can also undergo a conformational change
and bind to a second copy of Orf9b to form the homodimer. The Orf9b homodimer is in
equilibrium between free Orf9b monomers and Orf9b monomers bound to Tom70. Upon lipid
binding, the homodimer is tightly stabilized and slowly releases monomeric Orf9b.

151

CJ measured key rate constants that constrained a model to show how lipid binding shifts Orf9b’s equilibrium towards the homodimer, preventing Tom70 interaction. The lipid-free form comes to equilibrium in seconds, whereas the lipid-bound form takes hours-days!

A. I. Mathematical model overlay to FP competition kinetic assay using the lipid-bound
Orf9b homodimer as the competitor.
II. Mathematical model overlay to FP competition kinetic assay using the apo-Orf9b
homodimer as the competitor.

100

Here, we developed a quantitative biophysical model to describe the coupled equilibria of Orf9b dimerization and Tom70 binding, revealing how lipid binding tunes the system.

ODE for Tom70/Orf9b interactions:
1. d[Orf9b-FITC]/dt = -k1[Orf9b-FITC][Tom70]+k2[Orf9b-FITC:Tom70]
2. d[Orf9b β-monomer]/dt = 2(-k3[Orf9b β-monomer]^2+k4[Orf9b dimer])-kα[Orf9b β-
monomer]^2+kβ[Orf9b α-monomer]
3. d[Orf9b dimer]/dt = k3[Orf9b β-monomer]^2-k4[Orf9b dimer]
4. d[Tom70]/dt = -k1[Orf9b-FITC][Tom70]-k5[Orf9b α-monomer][T]+k6[Orf9b α-
monomer:Tom70]+k2[PT]
5. d[Orf9b α-monomer:Tom70]/dt = k5[Orf9b α-monomer][Tom70]-k6[Orf9b α-
monomer:Tom70]
6. d[Orf9b α-monomer]/dt = -k5[Orf9b α monomer][Tom70]-kβ[Orf9b α-monomer]+k6[Orf9b
α-monomer:Tom70]+kα[Orf9b β-monomer]^2

111

James Fraser @fraserlab.com · 05/02/2025

We show that Orf9b exists in a dynamic equilibrium between a homodimeric and monomeric state. The monomer binds to the mitochondrial protein Tom70, suppressing immune responses. Lipid binding to Orf9b stabilizes the homodimer, acting as a switch to slow down Tom70 engagement.

From left to right:

A large protein complex with several helices (colored in blue and purple) representing Orf9b bound to Tom70.
Orf9b detaching from Tom70 and undergoing a β-sheet to α-helix transition, shown in purple.
The α-helical form of Orf9b transitioning to a β-sheet-rich homodimer.
The final step showing the dissociation of the homodimer into individual monomers.
Arrows indicate equilibrium transitions between states, with labels highlighting key conformational changes: "β-sheet to α-helix" and "Dissociation".

121

New acknowledgement of @elife.bsky.social lack of Impact Factor in proofs... see response!

The image contains a message from "Exeter Pre-editing" regarding an update about the indexing of the Version of Record for eLife content. The main points are:

eLife content will be widely indexed, including by PubMed and OpenAlex.
However, starting June 2025, eLife content will not receive an impact factor.
A link is provided for more details: https://elifesciences.org/inside-elife/ae620829.
The recipient is asked to confirm awareness of the change and whether they are happy to continue.
The response box shows "Yes! Death to Impact Factor!" with the "Mark as resolved" checkbox selected.

050

James Fraser @fraserlab.com · 04/02/2025

Continuing the ☘️🍪▶️📄. Yesterday I was EXTREMELY disappointed to only receive 6 preprints (🧵) for 18 Caramel deLites. We have "preprint hating cookie thieves" (great name for a rock band!) on the 4th floor! I'm trying again - this time with the innovation of Scotch tape attached to the cookie box!

2110

James Fraser @fraserlab.com · 03/02/2025

Streptogramins are promising antibiotics for treating resistant bacterial infections, but their efficacy is threatened by Vat enzymes, which inactivate streptogramins by acetylation (blocking a key interaction with the ribosome). So we set out to block the blocker!

A structural illustration of the VatD trimer (left) showing its ribbon model with chains in gray, pink, and cyan. Acetyl-CoA (orange sticks) is bound in its pocket. Zoomed-in views (right) depict molecular interactions: the central panel shows Acetyl-CoA interacting with catalytic residues H82 and T88, while the rightmost panel highlights the binding of the antibiotic VM1 in the peptidyl transferase center (PTC) and nascent peptide exit tunnel (NPET), with key contacts involving Guanosine 2505 and a critical hydroxyl group (OH).

120

James Fraser @fraserlab.com · 03/02/2025

Our latest work on tackling streptogramin resistance in collaboration with the Seiple lab is now up as a preprint. We (led by Pooja Asthana) targeted a resistance enzyme, VatD, with fragment-based drug discovery! www.biorxiv.org/content/10.1...

An annotated molecular surface representation of the VatD trimer highlighting three fragment binding sites: the Active site (pink), the Acetyl-CoA (Ac-CoA) site (blue), and Accessory sites (yellow). Fragment-bound molecules are visualized as stick models at their respective binding locations, showing the spatial arrangement of the trimer interface.

1237

James Fraser @fraserlab.com · 03/02/2025

New strategy for discovering preprints. Cookies to Offline RSS feed! 🍪🍀

A sign saying that if you want a cookie you have to leave a preprint print out.

1181

Next, we took inspiration from more polar fragments binding lower in the original binding site and designed KABLE (Kemp eliminase ALBE). With a few additional mutations from a single round of screening it has a catalytic efficiency (600,000 M⁻¹s⁻¹) on par with the natural enzymes.

A diagram comparing the crystal structures of ABLE-fragment complexes to the design structure of the KABLE1-6NBT complex.

Left panel: The carboxylate-imidazole interaction in the crystal structures of ABLE bound to fragments is depicted. His49 is highlighted as a key residue, interacting with carboxylate-containing fragments.
Right panel: The design structure of KABLE1-6NBT is shown, where His49 is replaced with Asp49 to facilitate a strong 6NBT-carboxylate interaction. Additional key residues, Gln75 and Tyr75, are labeled and illustrated forming a hydrogen-bond network with 6NBT.
The transition highlights the rational design process of converting ABLE's binding site into a functional active site for catalyzing the Kemp elimination reaction.

100

Yuda Chen noticed that some of the fragments resembled coumarins. So they redesigned the binding site and created FABLE (Fluorescent ABLE) —a protein that binds turn-on fluorophores. ABLE->FABLE generalist to a specialist sensor with 100-fold turn-on fluorescence enhancement!

A stepwise illustration of the transformation of the protein ABLE into FABLE through fragment screening and sequence design.

Left (ABLE-Apixaban): The initial protein, ABLE, is shown bound to the drug apixaban, with the binding pocket highlighted and apixaban's structure displayed.
Second (ABLE-7-hydroxycoumarin): After fragment screening, ABLE binds the fragment 7-hydroxycoumarin, with its chemical structure shown above the pocket, demonstrating the potential to adapt to new ligands.
Third (ABLE-Cou485): Sequence design modifies the binding pocket to accommodate Cou485, a larger fluorophore. The structure of Cou485 is shown above the panel.
Right (FABLE-Cou485): The final designed protein, FABLE, specifically binds Cou485. The optimized binding pocket is highlighted, showing a tight fit for the fluorophore.
This progression illustrates the use of fragment screening and rational design to transform a generalist binder into a specialist for a turn-on fluorophore.

Lena Bergmann and Galen Correy used X-ray crystallography as a binding assay. ABLE turned out to be a generalist, binding 43 fragments in two distinct conformations. (side note, Nick and Justin Biel started this as COVID hit... but it paused until Lena picked it back up!)

$A diagram showcasing fragment screening and binding site characterization in a helical protein scaffold. Left panel: A single fragment bound in the B site of the protein is displayed, emphasizing the structural arrangement of the binding pocket. Center: A funnel graphic depicts the fragment screening workflow: 320 fragments screened, 242 processed (high-quality diffraction datasets collected), 43 hits identified, resulting in a hit rate of 18%. Right panel: A superimposed visualization of multiple fragments binding within the protein's B site, demonstrating the diversity of fragment binding poses. The fragments are color-coded, with the binding pocket highlighted.$

James Fraser @fraserlab.com · 23/01/2025

Thrilled to take a break from doom and gloom to share our latest work collaborating with the DeGrado lab! We took a de novo designed protein, screened it for ligand binding using X-rays, and used the hits to evolve two wildly different functions: fluorescent turn-on and Kemp elimination catalysis. 🧵

A diagram illustrating the evolution of protein functionality through sequence and chemical space.

ABLE (bottom left): Depicted as a helical bundle binding the drug apixaban, labeled as a "specialist" in binding apixaban. Its binding site is shown as a filled cavity with the drug's structure.
ABLE as a generalist (center): After fragment screening, ABLE evolves to weakly bind various small molecule fragments, shown in a diverse array within the cavity. This transition highlights the exploration of chemical space.
FABLE (top right): Derived from fragment-inspired design, FABLE binds turn-on fluorophores. The binding site is reconfigured to fit fluorophores like Cou485, with their chemical structures displayed.
KABLE (bottom right): Another fragment-inspired evolution turns ABLE into KABLE, a Kemp eliminase. The structure of the substrate (5-nitrobenzisoxazole) is shown, with the binding site adapted for catalysis.
The axes represent "Sequence Space" (vertical) and "Chemical Space" (horizontal), emphasizing how the protein evolves through these dimensions to achieve specialized functions.
This schematic encapsulates the journey of ABLE from a specific binder to a generalist and then into two specialized proteins with distinct functionalities.

38320

I've had such a lucky life in science. From time to time my mind drifts and I think about my two greatest mentors passing away too young. It stops me cold in my tracks. I've been staring at this picture for 15 minutes...

2241

James Fraser @fraserlab.com · 26/11/2024

Treating myself to a visit to the synchrotron for Thanksgiving with Galen and Julius - Photons on! 💎💫

2362

James Fraser @fraserlab.com · 18/11/2024

I recall this used to be a problem for Science too - but they have a good solution (figure legend on right, scrollable!) @richardsever.bsky.social

James Fraser @fraserlab.com · 18/11/2024

Haven't complained about this in a while, but now that @biorxivpreprint.bsky.social has such a great full text view for web... I HATE how clicking on the image brings up a magnified image that gets covered dynamically with the figure legend when you mouse over it... SO ANNOYING!!!!

image from biorxiv covered by figure legend!

1120

James Fraser @fraserlab.com · 16/11/2024

Just added the @bsky.app logo and links to our lab website members page (fraserlab.com/members) - these types of links are a great feature (along with the timeline, cartoon, and easter 🥚 pictures)... thanks to @tomo.science for providing the foundation - (to copy: fraserlab.com/2020/05/03/C...)

1579

James Fraser @fraserlab.com · 13/11/2024

Great drinks with Gira and Damian too!!!

020

James Fraser @fraserlab.com · 01/11/2023

Lab Halloween Costumes - (imagine the minus sign)...

1141

James Fraser @fraserlab.com · 04/10/2023

I see this all the time - you don't need to wait for a journal to ask you to review - JUST REVIEW THE PREPRINT and post your review! We do it all the time: fraserlab.com/reviews The majority of the time, a journal asks us to include the review in their process AFTER we post!

0137

James Fraser @fraserlab.com · 22/09/2023

these ones? already there