How to design proteins ?

Don't let the simplicity of this question fool you! Indeed, protein design hides multiple fundamental questions - which essentially shows how little we know about these very well-known molecules. well-known molecules.

Roughly speaking, two approaches are fighting: i) machine learning takes natural protein examples to design new ones, and ii) physics explicitly models the details hidden in protein atoms! I followed the second option, where in collaboration with experimentalists, we not only designed proteins for molecular structure stability but also new catalytic proteins (enzymes). The methods we worked on are implemented in the Proteus Software developed in Thomas Simonson's lab.

One gene can hide another one !

There is a natural process by which one gene can hide another gene — these genes are called overlapping genes. This sounds like magic, but it's not! This is simply due to two features of the genetic code: DNA is read by codons (triplets of nucleotides), and the genetic code is redundant.

Overlapping genes have multiple implications in evolutionary theories but not only. We devised an algorithm that can somehow "morph" two genes such that in the end both are intricated (overlap). You could only diferenciate the first gene from the second by reading the genetic sequence in a special way. Bluffing, right ?

After a Master in Bioinformatics at Paris-Saclay University (2015-2017), I started a PhD in Biology at Ecole Polytechnique (Institut Polytechnique de Paris). From 2017 to 2020, I worked on Computational Protein Design under the supervision of Pr. Thomas Simonson in the Structural Biology of the Cell laboratory. Next, I started a postdoc position at the Max Planck Institute for Mathematics in Sciences (Leipzig, Germany), from 2020 to 2022.