Our mission is to develop approaches to study protein structure and interactions in cells and make them universally accessible and useful.

Berlin, Germany and Edinburgh, Scotland
Joined August 2019
At Rappsilber Lab, we believe in innovation not only in science but also in how we care for our planet. 📷 By aligning with the My Green Lab standards, we’re committed to reducing our environmental footprint and creating a “greener” future for research and development.
1
192
The AlphaLink2 paper is now online! It’s joint work with @rbolab and @JorgStulke. AlphaLink2 integrates crosslinking MS data into AlphaFold-Multimer. AlphaLink2 really shines for challenging targets like antigen-antibody complexes. doi.org/10.1038/s41467-024-5…
32
1
86
7,308
We got an ERC synergy grant to work on and understand with Julia Mahamid (EMBL), Jan Kosinsky (EMBL), and Rachel Green (Johns Hopkins University, USA) ribosomes inside their cellular context and develop tools to study molecular processes inside cells at atomic resolution.
8
8
5
147
19,320
With AlphaLink2 we can now start visualizing protein-protein interactions inside cells at pseudo-atomic resolution by leveraging the very scarce data of whole-cell crosslinking MS. We thank @deepmind and Uni-Fold (@guolin_ke, @LiZiyao11) for providing excellent open-source tools
2
16
1,676
As a final note, sulfo-SDA crosslinks allow AlphaLink2 to position the viral protein within the Cullin4 complex's experimental density. (6/7)
1
4
6
1,658
The DockQ score would have been the same if only a single crosslink had been used. (5/7)
1
1
6
926
A validation of AlphaLink2 was conducted on real data from Bacillus subtilis cells crosslinked in situ with DSSO (O'Reilly et al.) and we demonstrated that even a single crosslink could significantly improve prediction accuracy. (4/7)
1
3
8
1,095
In this study, we demonstrate that crosslinking MS data significantly enhances the modelling of antibody-antigen targets. (3/7)
1
2
4
1,135
In collaboration with the @rbolab, we are pleased to present AlphaLink2 - a modified version of AlphaFold-Multimer that is capable of directly incorporating experimental residue-residue contact data from crosslinking MS (and potentially other techniques as well). (1/7)
3
76
2
241
32,114
The AlphaLink-Multimer manuscript is also in progress. We are pleased with the results. Enjoy a sneak peek at what's to come.
6
49
6,507
AlphaLink can represent crosslinking MS restraints as distograms which generalise to arbitrary distance restraints and allows mixing different types of (experimental) distance restraints. We thank @deepmind, @MoAlQuraishi and OpenFold for providing such excellent tools!
1
13
914
A few crosslinking MS restraints could act as “anchor points” to predict different conformational states of the kinase Cdk2 or fold-switching proteins. This has the potential to open up predictions of in-cell conformations of proteins from sparse restraints. (8/9)
1
1
13
1,022
One very interesting aspect of AlphaLink is the potential to “bias” structure predictions towards conformations observed under experimental conditions. (7/9)
1
1
9
597
The dataset included in-cell distance restraints on several membrane proteins, highlighting the potential of this novel technology. (6/9)
1
1
6
681
Therese Dau and Andrea Graziadei then produced and analysed a crosslinking MS dataset of the E. coli outer membrane using the “zero-length” crosslinker photo-leucine, which is directly incorporated into proteins. (5/9)
1
1
9
766
Kolja fine-tuned the network weights using simulated crosslinking MS data. AlphaLink leverages this experimental data to improve the predictions of OpenFold in the absence of large amounts of multiple sequence alignment information. (4/9)
1
1
6
720
Kolja Stahl used the excellent OpenFold from @MoAlQuraishi as a basis and then introduced biasing with experimental data by adding crosslinking MS restraints directly to the pair representation. (3/9)
1
1
7
777