Stockholm university

Samuel FloresDirector for Research School in Bioinformatics

About me

 

I work on macromolecular structure and dynamics, mostly from the computational side. I wrote MacroMoleculeBuilder (MMB), a multiresolution modeling tool for building 3D structural and dynamical models of macromolecules. It gives you explicit control over the degrees of freedom, forces, constraints, and molecules in your system. Published applications include predicting the effect of mutations on protein-protein interactions, in silico affinity maturation, RNA folding from biochemical data, homology modeling, flexible fitting into electron density maps, morphing, and fast prototype simulations of large complexes. Right now we are applying deep learning techniques to RNA structure prediction. We are also packaging viral DNA genomes. I have also written a few web servers, to design protein-protein interactions, for morphing, and for flexibility prediction.

MedBioInfo, The Swedish National Graduate School in Medical Bioinformatics

I am the Dean of Students (Studierektor) of MedBioInfo. We provide courses, an annual meeting, mentorship, and networking to PhD students in Bioinformatics at any Swedish university. If you are in your first year of a Bioinformatics program, you are encouraged to apply, read more at http://www.medbioinfo.se/ .   

 

Teaching

Teaching

I contribute some lectures and am doing pedagogical development for our course in Bioinformatics (KB7016/KB7021). 

Research

 

Current Research

What do we do once the world's genomes have been sequenced? There is still the question of what these molecules look like once they are transcribed into RNA (we work on this) and from there translated to protein (where Google's AlphaFold has made great progress) and what they do in the cell. 

 

In Silico design of biologics

I have written a few papers on predicting the change in protein-protein interaction energy (ΔΔG) resulting from mutations. I want to develop the capability to increase affinity in silico, and as a second priority, alter specificity. Recently I have developed a method which asks the user for a the Protein Data Bank (PDB) ID of a protein-protein complex, the chains in each of two interacting subunits, and a mutation to test. It not only computes ΔΔG using that structure (using the FoldX potential) but also searches the PDB for additional, sufficiently-similar structures on which to repeat the calculation. It then averages over the results of all such calculation, thus improving precision and (importantly) positive predictive value -- the fraction of mutations which are predicted to improve affinity, which successfully do so. You can try this out on our easy to use web server, http://biodesign.scilifelab.se/  .

RNA

The flexibility of large RNAs presents a significant challenge to understanding large complexes such as the ribosome, and to designing complexes for nanotechnology. Previously,  MMB has successfully folded tRNA and the P4/P6 domain of the Group I Intron, threaded the Azoarcus Group I Intron, explained the role of disease associated mutations in telomerase, and modeled all-atoms trajectories of ribosomal translocation. The latter drove development of very fast morphing and low-resolution (e.g. cryoem) density map fitting methods. We also written about GTP hydrolysis on EF-G and ribosomal frameshifting. Current projects include deep learning for RNA structure prediction, and viral DNA genome packaging.

Current MS students:

Kaouther Harbi

Floyd Jaggy

Emilio Skarwan