(1) Background: Inter-tumour heterogeneity is one of cancer’s most fundamental features. Patient stratification based on drug response prediction is hence needed for effective anti-cancer therapy. However, single-gene markers of response are rare and/or may fail to achieve a significant impact in the clinic. Machine Learning (ML) is emerging as a particularly promising complementary approach to precision oncology. (2) Methods: Here we leverage comprehensive Patient-Derived Xenograft (PDX) pharmacogenomic data sets with dimensionality-reducing ML algorithms with this purpose. (3) Results: Combining multiple gene alterations via ML leads to better discrimination between sensitive and resistant PDXs in 19 of the 26 analysed cases. Highly predictive ML models employing concise gene lists were found for three cases: paclitaxel (breast cancer), binimetinib (breast cancer) and cetuximab (colorectal cancer). Interestingly, each of these multi-gene ML models identifies some treatment-responsive PDXs not harbouring the best actionable mutation for that case. Thus, ML multi-gene predictors generally have much fewer false negatives than the corresponding single-gene marker. (4) Conclusions: As PDXs often recapitulate clinical outcomes, these results suggest that many more patients could benefit from precision oncology if ML algorithms were also applied to existing clinical pharmacogenomics data, especially those algorithms generating classifiers combining data-selected gene alterations.
9 January 2024 in Modeling Earth Systems and Environment
Burkholderia dolosa is a clinically important opportunistic pathogen in inpatients. Here we characterised an extensively drug-resistant and hypervirulent B. dolosa isolate from a patient hospitalised for stroke. Resistance to 41 antibiotics was tested with the agar disc diffusion, minimum inhibitory concentration, or broth microdilution method. The complete genome was assembled using short-reads and long-reads and the hybrid de novo assembly method. Allelic profiles obtained by multilocus sequence typing were analysed using the PubMLST database. Antibiotic-resistance and virulence genes were predicted in silico using public databases and the 'baargin' workflow. B. dolosa N149 phylogenetic relationships with all available B. dolosa strains and Burkholderia cepacia complex strains were analysed using the pangenome obtained with Roary. B. dolosa N149 displayed extensive resistance to 31 antibiotics and intermediate resistance to 4 antibiotics. The complete genome included three circular chromosomes (6 338 630 bp in total) and one plasmid (167 591 bp). Genotypic analysis revealed various gene clusters (acr, amr, amp, emr, ade, bla and tet) associated with resistance to 35 antibiotic classes. The major intrinsic resistance mechanisms were multidrug efflux pump alterations, inactivation and reduced permeability of targeted antibiotics. Moreover, 91 virulence genes (encoding proteins involved in adherence, formation of capsule, biofilm and colony, motility, phagocytosis inhibition, secretion systems, protease secretion, transmission and quorum sensing) were identified. B. dolosa N149 was assigned to a novel sequence type (ST2237) and formed a mono-phylogenetic clade separated from other B. dolosa strains. This study provided insights into the antimicrobial resistance and virulence mechanisms of B. dolosa.
8 May 2025 in Journal of Global Antimicrobial Resistance
Helicobacter pylori (H. pylori) is a clinically important pathogen associated with gastric diseases. Here, we characterized the genome of multidrug-resistant H. pylori strains of novel sequence types, which were recovered from Vietnamese patients with gastritis or a stomach ulcer. Phenotypic-antibiotic susceptibility testing was performed against amoxicillin, clarithromycin, metronidazol, tetracycline, and levofloxacin using an E-test. The whole genome sequence of three H. pylori strains was de novo assembled, followed by in silico analysis of multilocus sequence typing (MLST), core-genome based phylogeny, genetic determinants associated with virulence, biofilm formation, and antibiotic-resistance. The genome sequence of H. pylori strains exhibited a high similarity with the average nucleotide identity (ANI) values of 98.5% to 99.2%, carried 5 to 7 pathogenicity islands, and 3 to 6 mobilomes. The MLST profile of strains revealed novel sequence types ST4511, ST4512, and ST4513. Core-genome based phylogeny exhibited that the three H. pylori strains belong to the Asian genotype. These strains possessed 128 to 131 virulence genes, including toxin-encoding genes cagA and vacA. Double amoxicillin-resistant mutations on pbp1 and pbp2, or a mutation on pbp3, triple clarithromycin-resistant mutations on 23S rRNA gene and a levofloxacin-resistant mutation on gyrA were detected in antibiotic-resistant strains. Mutations on rdxA were detected in both metronidazole-resistant and -sensitive strains, whereas frxA mutations were detected in only one metronidazole-sensitive strain. Finally, a rifamycin-resistant mutation in rpoB was also detected. This study provides insights into the genome of multidrug-resistant H. pylori strains of a novel sequence type circulating in Vietnam for future investigations of its pathobiology and spread through human populations.
Prédiction de la réponse aux traitements in vivo de tumeurs basées sur le profil moléculaire des tumeurs par apprentissage automatique: Ces dernières années, les thérapies ciblées pour le traitement du cancer, ont été introduites. Cependant, un médicament fonctionnant chez un patient peut ne pas fonctionner chez un autre. Pour éviter l'administration de traitements inefficaces, des méthodes capables de prédire les patients qui répondront à un médicament donné doivent être mises au point.Il n'est actuellement pas possible de prédire l'efficacité de la grande majorité des médicaments anticancéreux. L’apprentissage automatique (AA) est un outil particulièrement prometteur pour la médecine personnalisée. L’AA est un champ d’étude de l'intelligence artificielle ; elle concerne la mise au point et l'application d'algorithmes informatiques qui s'améliorent avec l'expérience. Dans ce cas, l'algorithme d’AA apprendra à faire la distinction entre les tumeurs sensibles et résistantes en fonction de plusieurs gènes au lieu d'un seul gène. Cette étude se concentre sur l'application de différentes approches de l’AA pour prédire la réponse à des médicaments anticancéreux des tumeurs et générer des modèles précis, biologiquement pertinentes et faciles à expliquer.