Text-mining-based feature selection for anticancer drug response prediction.

Motivation: Predicting anticancer treatment response from baseline genomic data is a critical obstacle in personalized medicine. Machine learning methods are commonly used for predicting drug response from gene expression data. In the process of constructing these machine learning models, one of the most significant challenges is identifying appropriate features among a massive number of genes. Results: In this study, we utilize features (genes) extracted using the text-mining of scientific literatures. Using two independent cancer pharmacogenomic datasets, we demonstrate that text-mining-based features outperform traditional feature selection techniques in machine learning tasks. In addition, our analysis reveals that text-mining feature-based machine learning models trained on in vitro data also perform well when predicting the response of in vivo cancer models. Our results demonstrate that text-mining-based feature selection is an easy to implement approach that is suitable for building machine learning models for anticancer drug response prediction. Availability and implementation: https://github.com/merlab/text_features.

Characterization of the diversity of type IV secretion system-encoding plasmids in Acinetobacter.

The multi-drug resistant pathogen Acinetobacter baumannii has gained global attention as an important clinical challenge. Owing to its ability to survive on surfaces, its capacity for horizontal gene transfer, and its resistance to front-line antibiotics, A. baumannii has established itself as a successful pathogen. Bacterial conjugation is a central mechanism for pathogen evolution. The epidemic multidrug-resistant A. baumannii ACICU harbours a plasmid encoding a Type IV Secretion System (T4SS) with homology to the E. coli F-plasmid, and plasmids with homologous gene clusters have been identified in several A. baumannii sequence types. However the genetic and host strain diversity, global distribution, and functional ability of this group of plasmids is not fully understood. Using systematic analysis, we show that pACICU2 belongs to a group of almost 120 T4SS-encoding plasmids within four different species of Acinetobacter and one strain of Klebsiella pneumoniae from human and environmental origin, and globally distributed across 20 countries spanning 4 continents. Genetic diversity was observed both outside and within the T4SS-encoding cluster, and 47% of plasmids harboured resistance determinants, with two plasmids harbouring eleven. Conjugation studies with an extensively drug-resistant (XDR) strain showed that the XDR plasmid could be successfully transferred to a more divergent A. baumanii, and transconjugants exhibited the resistance phenotype of the plasmid. Collectively, this demonstrates that these T4SS-encoding plasmids are globally distributed and more widespread among Acinetobacter than previously thought, and that they represent an important potential reservoir for future clinical concern.

Large-scale phenogenomic analysis of human cancers uncovers frequent alterations affecting SMC5/6 complex components in breast cancer.

Cancer cells often experience large-scale alterations in genome architecture because of DNA damage and replication stress. Whether mutations in core regulators of chromosome structure can also lead to cancer-promoting loss in genome stability is not fully understood. To address this question, we conducted a systematic analysis of mutations affecting a global regulator of chromosome biology -the SMC5/6 complex- in cancer genomics cohorts. Analysis of 64 959 cancer samples spanning 144 tissue types and 199 different cancer genome studies revealed that the SMC5/6 complex is frequently altered in breast cancer patients. Patient-derived mutations targeting this complex associate with strong phenotypic outcomes such as loss of ploidy control and reduced overall survival. Remarkably, the phenotypic impact of several patient mutations can be observed in a heterozygous context, hence providing an explanation for a prominent role of SMC5/6 mutations in breast cancer pathogenesis. Overall, our findings suggest that genes encoding global effectors of chromosome architecture can act as key contributors to cancer development in humans.