Selecting a single model or combining multiple models for microarray-based classifier development?--a comparative analysis based on large and diverse datasets generated from the MAQC-II project

Transcriptome modulation by hydrocortisone in severe burn shock: ancillary analysis of a prospective randomized trial

Background

Despite shortening vasopressor use in shock, hydrocortisone administration remains controversial, with potential harm to the immune system. Few studies have assessed the impact of hydrocortisone on the transcriptional response in shock, and we are lacking data on burn shock. Our objective was to assess the hydrocortisone-induced transcriptional modulation in severe burn shock, particularly modulation of the immune response.

Methods

We collected whole blood samples during a randomized controlled trial assessing the efficacy of hydrocortisone administration in burn shock. Using whole genome microarrays, we first compared burn patients (n = 32) from the placebo group to healthy volunteers to describe the transcriptional modulation induced by burn shock over the first week. Then we compared burn patients randomized for either hydrocortisone administration or placebo, to assess hydrocortisone-induced modulation.

Results

Study groups were similar in terms of severity and major outcomes, but shock duration was significantly reduced in the hydrocortisone group. Many genes (n = 1687) were differentially expressed between burn patients and healthy volunteers, with 85% of them exhibiting a profound and persistent modulation over seven days. Interestingly, we showed that hydrocortisone enhanced the shock-associated repression of adaptive, but also innate immunity.

Conclusions

We found that the initial host response to burn shock encompasses wide and persistent modulation of gene expression, with profound modulation of pathways associated with metabolism and immunity. Importantly, hydrocortisone administration may worsen the immunosuppression associated with severe injury. These data should be taken into account in the risk ratio of hydrocortisone administration in patients with inflammatory shock.

Trial registration

ClinicalTrials.gov, NCT00149123. Registered on 6 September 2005.

Electronic supplementary material

The online version of this article (doi:10.1186/s13054-017-1743-9) contains supplementary material, which is available to authorized users.

Three general concepts to improve risk prediction: good data, wisdom of the crowd, recalibration

In today's information age, the necessary means exist for clinical risk prediction to capitalize on a multitude of data sources, increasing the potential for greater accuracy and improved patient care. Towards this objective, the Prostate Cancer DREAM Challenge posted comprehensive information from three clinical trials recording survival for patients with metastatic castration-resistant prostate cancer treated with first-line docetaxel. A subset of an independent clinical trial was used for interim evaluation of model submissions, providing critical feedback to participating teams for tailoring their models to the desired target. Final submitted models were evaluated and ranked on the independent clinical trial. Our team, called "A Bavarian Dream", utilized many of the common statistical methods for data dimension reduction and summarization during the trial. Three general modeling principles emerged that were deemed helpful for building accurate risk prediction tools and ending up among the winning teams of both sub-challenges. These principles included: first, good data, encompassing the collection of important variables and imputation of missing data; second, wisdom of the crowd, extending beyond the usual model ensemble notion to the inclusion of experts on specific risk ranges; and third, recalibration, entailing transfer learning to the target source. In this study, we illustrate the application and impact of these principles applied to data from the Prostate Cancer DREAM Challenge.

QSAR Models at the US FDA/NCTR

Quantitative structure-activity relationship (QSAR) has been used in the scientific research community for many decades and applied to drug discovery and development in the industry. QSAR technologies are advancing fast and attracting possible applications in regulatory science. To facilitate the development of reliable QSAR models, the FDA had invested a lot of efforts in constructing chemical databases with a variety of efficacy and safety endpoint data, as well as in the development of computational algorithms. In this chapter, we briefly describe some of the often used databases developed at the FDA such as EDKB (Endocrine Disruptor Knowledge Base), EADB (Estrogenic Activity Database), LTKB (Liver Toxicity Knowledge Base), and CERES (Chemical Evaluation and Risk Estimation System) and the technologies adopted by the agency such as Mold(2) program for calculation of a large and diverse set of molecular descriptors and decision forest algorithm for QSAR model development. We also summarize some QSAR models that have been developed for safety evaluation of the FDA-regulated products.

NETBAGs: a network-based clustering approach with gene signatures for cancer subtyping analysis

AIM

To evaluate gene signature and network-based approach for cancer subtyping and classification.

MATERIALS & METHODS

Here we introduced NETwork Based clustering Approach with Gene signatures (NETBAGs) algorithm, which clustered samples based on gene signatures and identified molecular markers based on their significantly expressed gene network profiles.

RESULTS

Applying NETBAGs to multiple independent breast cancer datasets, we demonstrated that the clustering results were highly associated with the clinical subtypes and clearly revealed the genomic diversity of breast cancer samples.

CONCLUSION

NETBAGs algorithm is able to classify samples by their genomic signatures into clinically significant phenotypes so that potential biomarkers can be identified. The approach may contribute to cancer research and clinical study of complex diseases.

CYCLoPs: A Comprehensive Database Constructed from Automated Analysis of Protein Abundance and Subcellular Localization Patterns in Saccharomyces cerevisiae

Changes in protein subcellular localization and abundance are central to biological regulation in eukaryotic cells. Quantitative measures of protein dynamics in vivo are therefore highly useful for elucidating specific regulatory pathways. Using a combinatorial approach of yeast synthetic genetic array technology, high-content screening, and machine learning classifiers, we developed an automated platform to characterize protein localization and abundance patterns from images of log phase cells from the open-reading frame−green fluorescent protein collection in the budding yeast, Saccharomyces cerevisiae. For each protein, we produced quantitative profiles of localization scores for 16 subcellular compartments at single-cell resolution to trace proteome-wide relocalization in conditions over time. We generated a collection of ∼300,000 micrographs, comprising more than 20 million cells and ∼9 billion quantitative measurements. The images depict the localization and abundance dynamics of more than 4000 proteins under two chemical treatments and in a selected mutant background. Here, we describe CYCLoPs (Collection of Yeast Cells Localization Patterns), a web database resource that provides a central platform for housing and analyzing our yeast proteome dynamics datasets at the single cell level. CYCLoPs version 1.0 is available at http://cyclops.ccbr.utoronto.ca. CYCLoPs will provide a valuable resource for the yeast and eukaryotic cell biology communities and will be updated as new experiments become available.

Gene expression profiles for predicting metastasis in breast cancer: a cross-study comparison of classification methods

Machine learning has increasingly been used with microarray gene expression data and for the development of classifiers using a variety of methods. However, method comparisons in cross-study datasets are very scarce. This study compares the performance of seven classification methods and the effect of voting for predicting metastasis outcome in breast cancer patients, in three situations: within the same dataset or across datasets on similar or dissimilar microarray platforms. Combining classification results from seven classifiers into one voting decision performed significantly better during internal validation as well as external validation in similar microarray platforms than the underlying classification methods. When validating between different microarray platforms, random forest, another voting-based method, proved to be the best performing method. We conclude that voting based classifiers provided an advantage with respect to classifying metastasis outcome in breast cancer patients.