We are proud to be part of @elixiruknode.org as the @unisouthampton.bsky.social node, and to share our expertise in structural and computational glycoscience 🧪, together with our OA resources thought it. Check out the announcement below for more information ⬇️ 🥳

a man in a suit and tie is giving a thank you sign . ALT: a man in a suit and tie is giving a thank you sign .

7/7 All links to ours and others OA resources and data for download are in the manuscript. As usual, any feedback is more than welcome. Happy glycosylating!

Vignettes showing in four stages the RF3-ReGlyco mini-binder design workflow to hEPO as in the demo. From the left-hand side, the 3D structure of hEPO is selected (yellow cartoons with a transparent molecular surface) as target protein. ReGlyco Ensemble is used to reconstruct the glycoprotein with the desired glycosylation. This process will output the inaccessible surface area due to the presence of glycans (dark red spheres) and the accessible surface area (yellow spheres) that can be targeted for binding, or ‘hotspots’. RF3 is used to design mini-binders. Generated designs are shown in grey cartoons. These designs are filtered by ReGlyco to exclude designs clashing with one or more glycosylation sites: predicted binders are shown in green cartoons (filter passed) and predicted non-binders (filter failed passed) in red cartoons. Mol* (https://molstar.org/) was used for structure rendering.

6/7 We made a colab notebook demo where users can design ‘mini binders’ against human erythropoietin (hEPO) by integrating GlycoShape with the RFdiffusion3 (RFD3) pipeline (doi.org/10.1101/2025...) from the Institute for Protein Design (IDP) colab.research.google.com/github/Ojas-...

kermit the frog is sitting on a railing wearing sneakers ALT: kermit the frog is sitting on a railing wearing sneakers

5/7 Adding glycosylation to the receptor in the AF3 prediction timed-out with no output after 24 hrs. We re-run those prediction recently with successful completion in 4 hrs for each design. AF3 is continuously improving, yet extremely time consuming (to date) vs AI+ReGlyco scheme

7th highest scoring binder id (submitted pose): Soft-Panda-Snow (149 aa). This pose failed the ReGlyco filter, clashing with glycan at N529. i. 7th highest scoring binder id: Soft-Panda-Snow (149 aa). Alternative pose obtained with AF3 (no glycans).

4/7 Enabling rotameric freedom (dunbrack.fccc.edu/lab/bbdep2010) refines the filter, yet still flagged 5 designs with irresolvable clashes in the Boltz-2 predicted structures. We rebuild all 5 complexes with the AF3 server (no glycans) and were able to obtain an alternative clear pose only for 1 ⬇️

Results of the Adaptyv Bio binder design competition without ReGlyco filtering are shown in the top bar. The middle bar shows the effect of ReGlyco filtering in flagging designs for clashes with one or more glycosylation sites (red) and pass (green). The bottom bar indicates the effect of the refinement with ReGlyco Rotamer, which introduces flexibility in the Asn sidechain of the aglycon. In yellow are confirmed binders flagged by Re-Glyco/Rotamer

3/7 We used GlycoShape ReGlyco and the new ReGlyco Rotamer tools to filter in-block the 1,201 results. This exercise flagged 11% of non-binders prior to experiment in approx 3 hrs on a dual-core CPU, a negligible computational overhead

a. Structure of the Nipah Virus Glycoprotein (NiV-G) homotetramer (aa 92-602) reconstructed from cryo-EM structures (PDB 7TXZ and 7TY0) bound to broadly neutralising antibody nAH1.3 Fabs30 (not shown). Static (single) glycans 3D structures (shown with sticks in blue) were reconstructed with GlycoShape ReGlyco17, where the choice of the glycan type was guided by glycoproteomics analysis36. The visible six glycan sites on chains A, B and D are mapped onto the structure, while glycans on chain C are not labelled for clarity. b. Structure of the glycosylated NiV-G recostructured with GlycoShape ReGlyco Ensemble29 using 150 frames from the MD trajectories of each selected glycan. Rendering effect is obtained with a “long exposure” filter to reflect the glycan dynamics. Sterically constrained glycans and rigid regions of the saccharides can be identified by the darker blue colour. c. Structure of the glycosylated NiV-G as in panel b. but with each one of the 150 glycan frames shown in dark blue, to illustrate the potentially occluded volume of the NiV-G protein surface.

2/7 To illustrate this point we chose to filter the results of a recent open competition launched by Adaptyv Bio for the design of binders to the heavily glycosylated Nipah virus glycoprotein (NiV-G). (proteinbase.com/competitions...) which is quite ‘furry’ when expressed in HEK293 cells

Most biologics are glycosylated and some of them heavily. In this “fresh off the press" #glycotime preprint we look into how a filter that accounts for glycosylation explicitly in 3D can help reduce lab costs and increase the efficiency of de novo binder design pipelines. Short 🧵 ⬇️ 1/7

From left: Prof Alba Silipo (UniNa, Federico II), Elisa Fadda, Dr Silvia D'Andrea, and Dr Trinidad Velasco-Torrijos, yellow table and amazing red velvet cake

Super congratulations to the Siglec (and many other stories) extraordinaire Dr. D'Andrea!!! 👩‍🎓🎉🥂🍾

A huge thank you to the fantastic examiners Alba Silipo (external) and Trinidad Velasco-Torrijos (internal) for your expert and in-depth examination, which made Silvia's viva a great day to remember 😎

PhD graduation photo of Dr Akash Satheesan. From the left Beatrice Tropea, Akash (showing his well deserved parchment), Ojas Singh and Silvia D'Andrea

Super congratulation to Dr Akash Satheesan from the glycoShape team, who graduated officially today with a PhD @maynoothuniversity.ie !! 😎🎓👏🥳🥂

The 🔺-red gown definitely suits you well Akash! 😍

Image representing Siglec-6 (surface cyan) digging through a membrane as if it was a field looking for GM1 (in yellow) the W127 is highlighted in pink as a plow

This work is also yet another example of how the chemistry and 3D context helps understanding recognition and binding. What can apparently look similar in 2D can be dramatically different in a 3D context where motions and dynamics are at play! Hope you'll enjoy reading the paper

a Binding of naked liposomes (liposomes that lack a ligand) to WT, W127A and Siglec-6 CHO cells using the cell assay (n = 4). Error bars correspond to standard deviation values. Figure generated using Adobe Illustrator. b IMS-CaR-nMS measurements for Siglec-6 (Fc) WT and mutants (each at 0.8 μM) with DMPC ND (5 μM) in 200 mM aqueous ammonium acetate solutions (pH 7.4); CID mass spectra (extracted form IMS heat maps) showing lipid ions released from (top) Siglec-6 WT, (middle) Siglec-6 R122A and (bottom) Siglec-6 W127A.

Finally, we demonstrated that Siglec-6 is a molecular precision tool, recognising not only the epitope with surgical precision, but also the environment where the epitope is found (see ⬇️). This helps rationalise the apparent redundancy of Siglecs as specific receptors to diff sialylated glycans

a Binding affinities (Kd, mM) of the ganglioside oligosaccharides GM1os, GM2os and GM3os for Siglec-6 (Fc) WT (light blue bars; n(GM1os) = 5, n(GM2os) = 3, n(GM3os) = 7) and W127A mutant (orange bars; n(GM1os) = 5, n(GM2os) = 3, n(GM3os) = 5) measured in aqueous ammonium acetate (200 mM, pH 6.8, 25 °C) by COIN-CaR-nMS. Bar plots generated with seaborn (https://seaborn.pydata.org/). Error bars correspond to standard deviationGL1 values in all panels. b Binding affinities (Kd, mM) of the ganglioside oligosaccharides GM1os, GM2os and GM3os for Siglec-6 (Fc) R122A mutant (green bars; n(GM1os) = 4, n(GM2os) = 4, n(GM3os) = 4) measured in aqueous ammonium acetate (200 mM, pH 6.8, 25 °C) by COIN-nMS. c–g IMS-CaR-nMS measurements performed in negative ion mode for 0.8 μM Siglec 6 (Fc) WT with 5 μM 10% GM1/DMPC ND in 200 mM aqueous ammonium acetate solution (pH 7.4). c IMS heat map (m/z versus IMS drift time tD) and d corresponding full mass spectrum extracted from IMS heat map using Driftscope. e IMS heat map and f corresponding extracted CID mass spectrum showing (glyco)lipid ions released from Siglec-6; ions with m/z 7000 ± 100 were isolated by quadrupole (isolation window highlighted in yellow in e followed by IMS and CID; a collision energy of 100 V was applied in the Transfer region. g IMS-CaR-nMS measurements for Siglec-6 (Fc) mutants (each at 0.8 μM) and 10% GM1/DMPC ND (5 μM) in 200 mM aqueous ammonium acetate solutions (pH 7.4); CID mass spectra (extracted form IMS heat maps) showing (glyco)lipid ions released from (top) Siglec-6 R122A and (bottom) Siglec-6 W127A. h 3D structure of the V-set domain of CD33 in complex with a sialoside analogue (PDB 7AW6). The protein is rendered in cyan as a surface, while the sialoside is rendered with sticks with C atoms in cyan, O in red and N in blue. SNFG symbols of the sialoside (left) and of the GM1 (right) are shown to indicate the position of the glucose at the reducing end.

Duong, Ling, Lara and John used native MS to measure the binding affinity of Siglec-6 to oligosaccharides, GM1os, GM2os and GM3os, and showed that in the absence of the bilayer, all three gangliosides are bound with the same affinity and in a Arg122 dependent manner🔥 😎

a Expression levels of the Siglec-6 WT and mutants obtained by flow cytometry (n = 4). Error bars correspond to standard deviation values in all panels. b Depiction and results of the cell assay used to assess the ability of Siglec-6 mutants to bind glycolipid liposomes using flow cytometry (n = 4). c Depiction and results of the ELISA approach used to assess the function of the W127A Siglec-6 Mutant (n = 4). Figures in panels a to c were generated using Adobe Illustrator and include only original elements. d 3D structures of the Ig V-set domains of MAG (blue cartoons; PDB 2ZG3), Siglec-6 (cyan cartoons; this work) and Siglec-11 (purple cartoons; AF-Q96RL6-F1). The residues in the membrane-facing loops that could potentially interact with the bilayer are labelled and highlighted with sticks. Molecular rendering done with Visual Molecular Dynamics55 (VMD; https://www.ks.uiuc.edu/Research/vmd/). e Sequence alignment of all human Siglecs performed with Clustal Omega58 (https://www.ebi.ac.uk/jdispatcher/msa/clustalo) shows that only Siglec-6 has a KW (or similar) combination of residues on the same loop, yet a number of other Siglecs, namely 5, 8, 10, 11, 14 and 16, have a combination of one aromatic and one positively charged residues across the two loops facing the cell membrane when binding gangliosides.

Eddie and Matt tested Siglec-6 binding on ganglioside-enriched liposomes and showed that the loss of K126 and W127 determine a complete loss of GM1 binding 🔥, and confirmed that the loss of the canonical Arg122 only decreases binding affinity, in perfect agreement with the 3D model 😎

Silvia's simulations showed that while all epitopes expose Sia for binding, Siglec-6 recognises and binds only GM1 because of a key interaction with the membrane through W127 and K126, which orientates the V-set domain to bind the Sia through Arg122 and the terminal Gal to the C-C' loop 😎

a 3D structure of Siglec-6 in complex with GM1 embedded in a lipid bilayer with composition distearoylphosphatidylcholine (DSPC) 60% and cholesterol 40%. In this representative snapshot from the MD trajectory, collected at 0.715 μs from the equilibrated MD ensemble, Arg 122 is engaged in a salt bridge with the Neu5Ac of GM1. The protein is represented with cyan cartoon rendering, the GM1 with sticks and C atoms in purple, O in red and N in blue. DSPC and cholesterol are rendered with semi-transparent sticks, with C atoms in grey, O in red, N in blue and P in yellow. Key residues are labelled with numbering corresponding to the human Siglec-6 (UniProtID O43699). b Structures of GM1, GM2 and GM3 represented with the SNFG nomenclature. Labels below each structure include the oligosaccharides GlyTouCan IDs. c Kernel Density Estimates (KDE) analysis of the tilt angle values measured though the 1.0 μs MD trajectories ran for isolated GM1 (purple) and GM3 (orange). KDE maxima are 31.32° and 41.83° measured for GM1 and GM3, respectively. d 3D structure of an isolated GM1 molecule (sticks with C atoms in purple, O in red and N in blue) embedded in one of the bilayer leaflets (sticks with all atoms in grey) used to represent the axes used to measure the tilt angle (θ) indicating the orientation relative to the bilayer of the Neu5Ac and thus its accessibility. e Close-up view on the Siglec-6 binding site obtained through a counterclockwise rotation of approximately 120° relative to the structure in a. Key residues are labelled, while the embedding of the Trp127 sidechain in the bilayer is highlighted by a surface rendering of the lipids. The terminal Gal of GM1 is highlighted with C atoms in yellow for his role in orienting the C-C’ loop in the bound complex, also shown in yellow. f Time evolution along the MD trajectory of the distance (Å) between the Arg122 and the Neu5Ac carboxylate group. Data points correspond to the largest distance value calculated between four pairs of…

Silvia designed and ran a 2 years worth of MD simulations to determine how recognition of a very small epitope like GM1 sticking out of the membrane could be recognised and bound specifically by Siglec-6, i.e only GM1, and not GM2 or GM3 which are all very similar...

Figure 1. Schematic representation of human Siglecs. Evolutionarily conserved Siglecs (Group 1) are shown on the left-hand side panel (pink box). The CD33-related Siglecs (Group 2) are shown on the right-hand side panel, (light blue box). The Siglecs extracellular domains are represented by a composition of Ig domains (C2-set) rendered as a solvent accessible surface, highlighted within green circles. The terminal Ig V-set binding domain is highlighted within an orange circle with a bound sialic acid in magenta (PDB 1OD9). Cytoplasmic domains are indicated with boxes and labelled according to the legend. Siglec-6 is highlighted within a yellow box as it is the focus of this work. All structural elements in this image were rendered with pymol (www.pymol.org) and incorporated in an original design based on Fig. 1 in ref. 2. In the insert on the top right-hand side, the conserved Arg is shown in the 3D structure of the Ig domain V-set of Siglec-6 (orange cartoons) bound to the NeuNAc-a(2-3)-Gal fragment (sticks with purple C atoms, red O atoms and blue N atoms) as an example.

Siglec-6 is one of 14 human Siglecs, all known to bind Sia with through a conserved (canonical) Arg. In earlier work Eddie, Matt et al (doi.org/10.1038/s414...) determined that in Siglec-6 the canonical Arg is dispensable, loss of R122 decreases but does not eliminate binding. How does this work?

a group of people are standing next to each other in a hallway and one of them is holding a book . ALT: a group of people are standing next to each other in a hallway and one of them is holding a book .

First, thank yous
1) to the very talented @silviadandrea.bsky.social for never giving up and seeing this through and @ojas-singh.bsky.social for phenomenal help with data analytics,
2) the 🇨🇦 team @siglecdude.bsky.social John Klassen and @glycocode.bsky.social, Eddie, Duong, and Ling 👏👏👏
⬇️🧵

Lectins are known to be low affinity binders with relatively broad glycan target preference. In this work live in @commsbio.nature.com we show how Siglecs change this paradigm, acting as molecular precision tools required to fine tune immune response 🤯 #glycotime 🧵⬇️

doi.org/10.1038/s420...

GLYCO-BUILD: an enzymatic pipeline for the synthesis of peptides carrying eukaryotic N-glycans - Nature Communications Rossi and colleagues assemble an in vitro enzymatic pipeline using enzymes from distinct domains of life to recapitulate eukaryotic N-glycosylation. This work advances the synthesis of bespoke glycope...

Phenomenal #glycotime by Lorenzo Rossi, @j-a-n-alexander.bsky.social, @asramirez.bsky.social and Kaspar Locher ⬇️ shows a pipeline for building THE glycoform you want with unprecedented yields!

Now you can design your glycoform with GlycoShape and make it with Glyco-BUILD 👏

doi.org/10.1038/s414...

As a bonus, because it's a new year we have updated the Mol* viewer with fullscreen and VR support 👍 happy glycosylating 🤓😍

Screenshot of the Advanced Settings menu in Re-Glyco Ensemble, where you can select 50 up to 500 structures (frames) from MD to represent the dynamics of the glycan on the glycoprotein you are rebuilding. You can calculate the Solvent Accessible Surface (SASA) the format of the output, i.e. standard PDB or GLYCAM for simulations with the Amber ff and from today the seed to guarantee reproducibility of the output. The seed is optional, if you like to keep the selection of conformers across the population clusters random, you can leave that space blank

This ⬇️ is screenshot of the Advanced Settings I used to produce the ensemble shown in the picture above,

Structure of a protein (cyan/teal) rendered as surface with 50 frames representing the dynamic ensemble of the glycans (white surface and sticks) depicted at the three sites (N93, N99 and N104) indicated by the 2D SNFG symbols and corresponding GlyTouCan IDs. Graphic rendering with VMD

Happy new year 2026 #glycotime! 🥳 We just introduced some useful options to Re-Glyco Ensemble.

In the Advanced Settings you can now select a seed (positive integer) to reproduce structural ensemble of your favourite glycoforms such as the one below that you will see soon in an upcoming paper,

elvis presley is dancing in a black and white photo with a group of men playing drums and guitars . ALT: elvis presley is dancing in a black and white photo with a group of men playing drums and guitars .

And a super special THANK YOU to Oracle for unparalleled support from the get-go through Oracle for Research (Rich Pitts and Mike Reilly 🙏) and now through OCI phenomenal support!

Stay tuned for a whole lot of 🔥 stuff coming up in the new year!

Screenshot of our current members of the eLab > Team tab, scroll down for all members who contributed to the work and now moved to their new careers

GlycoShape couldn't have happened without all the hard work, super skillzzz and dedication of our team, current (below) and previous members! 🤩🤩 EF is not the only cat, you can scroll over the photos to discover all others 😍 And key funding from @researchireland.ie (former SFI) FFP🙏 ⬇️🧵

We have used this approach to rebuild hyperglycosylated human EPO (or NESP shown below in the gif) in seconds 🏎️💨 as part of MSci and postgraduate summer schools structure glycoengineering for biologics workshop ⬇️🧵

screenshot of the selection of frames to include in the Re-Glyco Ensemble analysis. Here I selected 100 conformations for each site, but you can go as low as 50 (default) or 500. To do this open advanced settings

We have also a new and improved Re-Glyco Ensemble, where you can see through a multiframe (50 to 500 frames from MD) glycan structure view, statistics and corresponding SASA analysis how the protein structure shifts the conformational equilibrium of the glycans ⬇️🧵

screenshot of the result of the re-glycosylation of NESP with Re-Glyco, you can see the queuing information above the structure, below the Mol* structure GUI (green protein in cartoons with glycans represented by SNFG symbols) you find buttons to can download the re-glycosylated structure (left) and switch to Ensemble mode (purple) where you can add glycan dynamic info to the static view

We are not only continuing to grow our library, but also continuously perfecting and expanding our tools. Re-Glyco is leading the way 🤩, with improved computational efficiency 🏎️💨 and informative queuing widget and calculation process log ⬇️🧵

Screenshot of the "Database" front page indicating on the left the total number of glycans. You can search this database by filtering through your target mass range or by drawing the glycan with the Glycan Drawer you can find under "Draw" on the left of the search box, or by searching by glytoucan ID

To do that we built the largest OA 3D library of glycan structures worldwide 🤯, counting 882 unique glycans complemented by at least 3 conformers each to account for their dynamics, obtained from the analysis of multimicrosecond MD simulations doi.org/10.1038/s415... ⬇️🧵

World map with each blue dot represents a location of a glycoshape user. Data collected from launch to Nov 2025. land in white and water in light blue. The pie chart on the left indicates the countries where most users are based, first US, followed by China and UK third

Since Dec 2023 Glycoshape has progressively built a community of glycoengineers that continues to grow. These scientists recognise that to understand glycoproteins structure and function we need to take into consideration the glycans they have, with their micro and macro heterogeneity 🧵⬇️