Ghis­laine Gayraud and Miraine Davi­la Felipe, respec­tive­ly uni­ver­si­ty pro­fes­sor and senior lec­tur­er at UTC’s LMAC, are inter­est­ed in the sto­chas­tic mod­el­ling of com­plex real sys­tems. Irene Maf­fuc­ci, a senior lec­tur­er attached to GEC, spe­cialis­es in struc­tur­al bioin­for­mat­ics. The three sci­en­tists are par­tic­i­pat­ing in the Lyme dis­ease project led by Séver­ine Padiolleau.

What is the role played by math­e­mati­cians in this project? “The idea is to pro­pose a sto­chas­tic mod­el that gen­er­ates in sil­i­co new DNA sequences resem­bling (from a prob­a­bilis­tic point of view) those obtained in vit­ro by Séver­ine’s team. From this fam­i­ly of sequences gen­er­at­ed in sil­i­co, the next step is to select a few, known as ‘probes’, which appear to have good pair­ing capa­bil­i­ties with the pro­tein of inter­est. Their selec­tion con­sists of look­ing for those that are ” close”, in terms of math­e­mat­i­cal dis­tance, to those that Séver­ine and Irene’s team have labelled as suit­able can­di­dates from among all the sequences obtained in vit­ro by SELEX. The entire study process begins with the exper­i­men­tal part (SELEX), con­tin­ues with a math­e­mat­i­cal mod­el and tools, then back to the exper­i­men­tal part to val­i­date the sequences select­ed in sil­i­co,” explains Ghis­laine Gayraud.

What does “sequence” is mean”? “DNA is com­posed of four nitroge­nous bases: A, T, C, G. A sequence is the suc­ces­sion of these bases posi­tioned in a par­tic­u­lar order and over a giv­en length. The diver­si­ty of 10¹⁴ mol­e­cules men­tioned by Stéphane cor­re­sponds to 10¹⁴ dif­fer­ent sequences of length 40 at the ini­tial stage of SELEX. The aim of this pro­ce­dure is to iden­ti­fy poten­tial probes in vit­ro,” empha­sis­es Séver­ine Padiolleau.

How does the in vit­ro pro­ce­dure work? “Dur­ing this very time-con­sum­ing and com­plex process, there will be a num­ber of sequences capa­ble of recog­nis­ing the pro­tein of inter­est, but still few­er than all the pos­si­bil­i­ties. This is a major lim­i­ta­tion of SELEX,” explains Miraine Davi­la Felipe.

This is where math­e­mat­ics comes in. “Based on our in vit­ro results, Ghis­laine and Miraine will pro­pose a math­e­mat­i­cal mod­el to gen­er­ate in sil­i­co new sequences that are dif­fer­ent from ours but poten­tial­ly func­tion­al. Of course, one might won­der what the point of this is, since we already have a good can­di­date in vit­ro. In fact, tar­get­ing the same tar­get is not an end in itself. On the one hand, it increas­es the num­ber of probes avail­able. On the oth­er hand, this mod­el could be applied to tar­get oth­er pro­teins of inter­est with­out hav­ing to imple­ment the SELEX pro­ce­dure,” details Séver­ine Padiolleau.

Ghis­laine and Miraine, sup­port­ed by a post-doc, drew on an exist­ing fam­i­ly of mod­els, the Restrict­ed Boltz­mann Machines (RBM) fam­i­ly. ” The RBM is a two-lay­er graph­i­cal mod­el with an input lay­er con­sist­ing of the sequence and a hid­den lay­er that is sup­posed to take into account the 2D or, even bet­ter, the 3D struc­ture of the sequence when it folds. The pres­ence of this hid­den lay­er is impor­tant because the way the sequence folds is fun­da­men­tal to ensur­ing its pair­ing with the tar­get pro­tein,” adds Ghislaine.

Numerous thesis in support of the project

Work on Lyme dis­ease began in 2021 with Mick­aël Guérin’s the­sis. At UTC’s GEC Lab­o­ra­to­ry, it has con­tin­ued since 2024 with Hugo Da-Pon­te’s the­sis super­vised by Séver­ine Padi­ol­leau, Sel­ma Ben­guaouer’s the­sis co-super­vised by Irene Maf­fuc­ci and Bérangère Bihan-Avalle, and the sup­port of Pauline Trézel, a post-doc. At the LMAC, post-doc El Meh­di Issouani, who is sup­port­ing Miraine Davi­la Felipe and Ghis­laine Gayraud. Marc Shawky is cur­rent­ly co-super­vis­ing, with Flo­ri­an de Vuyst, Tere­sa Cia­vat­tini’s the­sis on the clas­si­fi­ca­tion of patient data in coop­er­a­tion with Mater Mis­eri­cor­diae Uni­ver­si­ty Hos­pi­tal, Dublin, Ire­land. Final­ly, a post-doc­tor­al researcher will be recruit­ed to devel­op machine learn­ing tools to assist clin­i­cians in their deci­sions to start a giv­en treatment.

MSD