Société Française de Bioinformatique

Mots-Clés

Protein structure/function Gene annotation Machine learning Comparative genomics Minimal bacteria Metabolic modelling

Description

Résumé du projet de stage:

Huge advances in microbiology have been made since the advent of genome sequencing, largely due to the prediction of metabolic “function” of a given microbe from the genes in its genome. These genes are traditionally annotated by sequence comparison to large databases of characterized genes. Despite ever-improving genome sequencing technology, many individual genes (and their respective proteins) remain unannotated by traditional sequence homology methods. These unannotated, or “hypothetical” genes, are referred to as “microbial dark matter”, and are estimated to comprise ~14% of all protein-coding genes (1). With the development of AI and large language models, a range of new methods have come available for structural and functional inference from protein sequences, such as AlphaFOLD (2), RoseTTAfold (3), and CLEAN (4). These methods can be used to shed light on the potential functions of genes/proteins classified as microbial dark matter.
Mollicutes are a group of small bacteria with reduced genomes (5-10x smaller than most classical bacteria) that cause disease in a wide range of hosts—both plant and animal. Many species of Mollicutes remain uncultivated or very difficult to grow in a laboratory environment (ex: Candidatus Phytoplasma vitis Flavescence dorée, a grapevine pathogen (5), and Spiroplasma citri, a citrus pathogen(6–8)). Their small genomes encode many functionally unannotated protein-coding genes, whose characterization could help increase cultivability.
Our team has selected 36 mollicutes genomes for comparative genomics analysis, and produced a genome-scale metabolic network (GSMN) from each genome. GMSNs represent every metabolic reaction that could occur in a given microbe—based on the functional annotation of its genome—and are used to predict growth of a microbe in a certain environment when provided certain nutrients. Because these networks directly depend on a genome’s functional annotation, improved annotation could drastically improve the applicability of GSMNs, which would ultimately help cultivation efforts.
The intern will be directly supervised by experts in bioinformatics, genomics, and metabolic modelling, in close collaboration with experts in Mollicutes biology and genomics.

Objectives of the internship include:

• Review and compare computational tools for annotation of unannotated genes/proteins
• Identify potential functions of hypothetical proteins in two highly curated genomes by implementing various machine learning algorithms
• Track and compare unannotated proteins across 36 mollicutes genomes to determine whether they are shared amongst cultivable/ uncultivable strains
• Assess the impact of these novel annotations on microbial metabolism and cultivability using GSMNs
• Summarize and present findings

Expected skills and profile

Expected:

• Proficiency in Python
• Good level in English (written and spoken)
• Interest in microbiology/ biochemistry

Appreciated:

• High-performance computing
• Bash programming language

Working language:

• English

We are seeking a student with one of the following profiles:

• Master’s degree in computational biology
• or a Master’s degree in computer science or artificial intelligence and an interest in biology

Montant des indemnités de stage: gratification au taux en vigueur.

Références

1. Pavlopoulos GA, Baltoumas FA, Liu S, Selvitopi O, Camargo AP, Nayfach S, Azad A, Roux S, Call L, Ivanova NN, Chen IM, Paez-Espino D, Karatzas E, Acinas SG, Ahlgren N, Attwood G, Baldrian P, Berry T, Bhatnagar JM, Bhaya D, Bidle KD, Blanchard JL, Boyd ES, Bowen JL, Bowman J, Brawley SH, Brodie EL, Brune A, Bryant DA, Buchan A, Cadillo-Quiroz H, Campbell BJ, Cavicchioli R, Chuckran PF, Coleman M, Crowe S, Colman DR, Currie CR, Dangl J, Delherbe N, Denef VJ, Dijkstra P, Distel DD, Eloe-Fadrosh E, Fisher K, Francis C, Garoutte A, Gaudin A, Gerwick L, Godoy-Vitorino F, Guerra P, Guo J, Habteselassie MY, Hallam SJ, Hatzenpichler R, Hentschel U, Hess M, Hirsch AM, Hug LA, Hultman J, Hunt DE, Huntemann M, Inskeep WP, James TY, Jansson J, Johnston ER, Kalyuzhnaya M, Kelly CN, Kelly RM, Klassen JL, Nüsslein K, Kostka JE, Lindow S, Lilleskov E, Lynes M, Mackelprang R, Martin FM, Mason OU, McKay RM, McMahon K, Mead DA, Medina M, Meredith LK, Mock T, Mohn WW, Moran MA, Murray A, Neufeld JD, Neumann R, Norton JM, Partida-Martinez LP, Pietrasiak N, Pelletier D, Reddy TBK, Reese BK, Reichart NJ, Reiss R, Saito MA, Schachtman DP, Seshadri R, Shade A, Sherman D, Simister R, Simon H, Stegen J, Stepanauskas R, Sullivan M, Sumner DY, Teeling H, Thamatrakoln K, Treseder K, Tringe S, Vaishampayan P, Valentine DL, Waldo NB, Waldrop MP, Walsh DA, Ward DM, Wilkins M, Whitman T, Woolet J, Woyke T, Iliopoulos I, Konstantinidis K, Tiedje JM, Pett-Ridge J, Baker D, Visel A, Ouzounis CA, Ovchinnikov S, Buluç A, Kyrpides NC. 2023. Unraveling the functional dark matter through global metagenomics. Nature 2023 622:7983 622:594–602. DOI:10.1038/s41586-023-06583-7.
2. Jumper J, Evans R, Pritzel A, Green T, Figurnov M, Ronneberger O, Tunyasuvunakool K, Bates R, Žídek A, Potapenko A, Bridgland A, Meyer C, Kohl SAA, Ballard AJ, Cowie A, Romera-Paredes B, Nikolov S, Jain R, Adler J, Back T, Petersen S, Reiman D, Clancy E, Zielinski M, Steinegger M, Pacholska M, Berghammer T, Bodenstein S, Silver D, Vinyals O, Senior AW, Kavukcuoglu K, Kohli P, Hassabis D. 2021. Highly accurate protein structure prediction with AlphaFold. Nature 2021 596:7873 596:583–589. DOI:10.1038/s41586-021-03819-2.
3. Baek M, DiMaio F, Anishchenko I, Dauparas J, Ovchinnikov S, Lee GR, Wang J, Cong Q, Kinch LN, Schaeffer RD, Millán C, Park H, Adams C, Glassman CR, DeGiovanni A, Pereira JH, Rodrigues A V., Dijk AA van, Ebrecht AC, Opperman DJ, Sagmeister T, Buhlheller C, Pavkov-Keller T, Rathinaswamy MK, Dalwadi U, Yip CK, Burke JE, Garcia KC, Grishin N V., Adams PD, Read RJ, Baker D. 2021. Accurate prediction of protein structures and interactions using a three-track neural network. Science (1979) 373:871–876. DOI:10.1126/science.abj8754.
4. Yu T, Cui H, Li JC, Luo Y, Jiang G, Zhao H. 2023. Enzyme function prediction using contrastive learning. Science (1979) 379:1358–1363. DOI:10.1126/science.adf2465.
5. Saglio P, Lhospital M, Lafleche D. 1973. Spiroplasma citri gen. and sp. n.: a mycoplasma like organism associated with “Stubborn” disease of citrus. Int J Syst Bacteriol 23:191–204. DOI:10.1099/00207713-23-3-191.
6. Davis RE, Shao J, Zhao Y, Gasparich GE, Gaynor BJ, Donofrio N. 2017. Complete Genome Sequence of Spiroplasma citri Strain R8-A2T, Causal Agent of Stubborn Disease in Citrus Species. Genome Announc 5:e00206-17. DOI:10.1128/genomea.00206-17.
7. Saillard C, Carle P, Duret-Nurbel S, Henri R, Killiny N, Carrère S, Gouzy J, Bové JM, Renaudin J, Foissac X. 2008. The abundant extrachromosomal DNA content of the Spiroplasma citri GII3-3X genome. BMC Genomics 9:195. DOI:10.1186/1471-2164-9-195.
8. Debonneville C, Mandelli L, Brodard J, Groux R, Roquis D, Schumpp O. 2022. The Complete Genome of the “Flavescence Dorée” Phytoplasma Reveals Characteristics of Low Genome Plasticity. Biology (Basel) 11:953. DOI:10.3390/biology11070953/s1.