Mario Lauria

Strengths and limitations of microarray-based phenotype prediction: lessons learned from the IMPROVER Diagnostic Signature Challenge

MOTIVATION After more than a decade since microarrays were used to predict phenotype of biological samples, real-life applications for disease screening and identification of patients who would best benefit from treatment are still emerging. The interest of the scientific community in identifying best approaches to develop such prediction models was reaffirmed in a competition style international collaboration called IMPROVER Diagnostic Signature Challenge whose results we describe herein. RESULTS Fifty-four teams used public data to develop prediction models in four disease areas including multiple sclerosis, lung cancer, psoriasis and chronic obstructive pulmonary disease, and made predictions on blinded new data that we generated. Teams were scored using three metrics that captured various aspects of the quality of predictions, and best performers were awarded. This article presents the challenge results and introduces to the community the approaches of the best overall three performers, as well as an R package that implements the approach of the best overall team. The analyses of model performance data submitted in the challenge as well as additional simulations that we have performed revealed that (i) the quality of predictions depends more on the disease endpoint than on the particular approaches used in the challenge; (ii) the most important modeling factor (e.g. data preprocessing, feature selection and classifier type) is problem dependent; and (iii) for optimal results datasets and methods have to be carefully matched. Biomedical factors such as the disease severity and confidence in diagnostic were found to be associated with the misclassification rates across the different teams. AVAILABILITY The lung cancer dataset is available from Gene Expression Omnibus (accession, GSE43580). The maPredictDSC R package implementing the approach of the best overall team is available at www.bioconductor.org or http://bioinformaticsprb.med.wayne.edu/.

Crowdsourced estimation of cognitive decline and resilience in Alzheimer's disease

Identifying accurate biomarkers of cognitive decline is essential for advancing early diagnosis and prevention therapies in Alzheimers disease. The Alzheimers disease DREAM Challenge was designed as a computational crowdsourced project to benchmark the current state‐of‐the‐art in predicting cognitive outcomes in Alzheimers disease based on high dimensional, publicly available genetic and structural imaging data. This meta‐analysis failed to identify a meaningful predictor developed from either data modality, suggesting that alternate approaches should be considered for prediction of cognitive performance.

Rank-based transcriptional signatures

We have developed a method for the definition and the analysis of gene expression signatures for diagnostic purposes. Our approach relies on construction of a reference map of transcriptional signatures, from both healthy controls and affected patients, using the respective mRNA or miRNA profiles. Subsequently, disease diagnosis can be performed by determining the relative map position of an individual’s transcriptional signature. Our approach addresses simultaneously the scarce repeatability issue and the high sensitivity of expression profiling methods to protocol variations, thereby providing a novel approach to RNA signature definition and analysis. Specifically, our method requires only that the relative position of RNA species be accurate in a ranking by value, not their absolute values. Furthermore, our method makes no assumptions on which RNA species must be included in the signature and, by considering a large subset (or even the whole set) of known RNAs, our approach can tolerate a moderate number of erroneous inversions in the ranking. The diagnostic power of our method has been convincingly demonstrated in an open scientific competition (sbv IMPROVER Diagnostic Signature Challenge), scoring second place overall, and first place in one sub-challenge. In addition, we report the application of our method to published miRNA expression profile data sets, quantifying its performance in terms of predictive capability and robustness to batch effects, compared with current state-of-the-art methods.

Systems biology meets -omic technologies: novel approaches to biomarker discovery and companion diagnostic development

The next generation of biomarkers and companion diagnostics will require the development of technologies capable of conjugating the advances in high-throughput techniques in biology with computational methods. Systems biology is poised to contribute through an integrated view, capturing the complexity of the system, both in terms of a collection of interacting molecular components and also in terms of multiple intersecting views. Following this system-centered view, novel approaches have been developed for the identification of signatures of both disease processes and drug modes of action with the promising perspectives of better diagnosis of disease and of the discovery of more efficacious and safe drugs. The application of systems biology to the development of companion diagnostics is very recent and to date a few pioneering steps have been made in this direction. In this review, we describe the ongoing studies and the potential developments in this area of research.

Diversity of key players in the microbial ecosystems of the human body

Coexisting bacteria form various microbial communities in human body parts. In these ecosystems they interact in various ways and the properties of the interaction network can be related to the stability and functional diversity of the local bacterial community. In this study, we analyze the interaction network among bacterial OTUs in 11 locations of the human body. These belong to two major groups. One is the digestive system and the other is the female genital tract. In each local ecosystem we determine the key species, both the ones being in key positions in the interaction network and the ones that dominate by frequency. Beyond identifying the key players and discussing their biological relevance, we also quantify and compare the properties of the 11 networks. The interaction networks of the female genital system and the digestive system show totally different architecture. Both the topological properties and the identity of the key groups differ. Key groups represent four phyla of prokaryotes. Some groups appear in key positions in several locations, while others are assigned only to a single body part. The key groups of the digestive and the genital tracts are totally different.

Systems view of adipogenesis via novel omics-driven and tissue-specific activity scoring of network functional modules.

The investigation of the complex processes involved in cellular differentiation must be based on unbiased, high throughput data processing methods to identify relevant biological pathways. A number of bioinformatics tools are available that can generate lists of pathways ranked by statistical significance (i.e. by p-value), while ideally it would be desirable to functionally score the pathways relative to each other or to other interacting parts of the system or process. We describe a new computational method (Network Activity Score Finder - NASFinder) to identify tissue-specific, omics-determined sub-networks and the connections with their upstream regulator receptors to obtain a systems view of the differentiation of human adipocytes. Adipogenesis of human SBGS pre-adipocyte cells in vitro was monitored with a transcriptomic data set comprising six time points (0, 6, 48, 96, 192, 384 hours). To elucidate the mechanisms of adipogenesis, NASFinder was used to perform time-point analysis by comparing each time point against the control (0 h) and time-lapse analysis by comparing each time point with the previous one. NASFinder identified the coordinated activity of seemingly unrelated processes between each comparison, providing the first systems view of adipogenesis in culture. NASFinder has been implemented into a web-based, freely available resource associated with novel, easy to read visualization of omics data sets and network modules.

SCUDO: a tool for signature-based clustering of expression profiles

SCUDO (Signature-based ClUstering for DiagnOstic purposes) is an online tool for the analysis of gene expression profiles for diagnostic and classification purposes. The tool is based on a new method for the clustering of profiles based on a subject-specific, as opposed to disease-specific, signature. Our approach relies on construction of a reference map of transcriptional signatures, from both healthy and affected subjects, derived from their respective mRNA or miRNA profiles. A diagnosis for a new individual can then be performed by determining the position of the individuals transcriptional signature on the map. The diagnostic power of our method has been convincingly demonstrated in an open scientific competition (SBV Improver Diagnostic Signature Challenge), scoring second place overall and first place in one of the sub-challenges.

Integration of transcriptomic and genomic data suggests candidate mechanisms for APOE4-mediated pathogenic action in Alzheimer’s disease

Among the genetic factors known to increase the risk of late onset Alzheimer’s diseases (AD), the presence of the apolipoproteine e4 (APOE4) allele has been recognized as the one with the strongest effect. However, despite decades of research, the pathogenic role of APOE4 in Alzheimer’s disease has not been clearly elucidated yet. In order to investigate the pathogenic action of APOE4, we applied a systems biology approach to the analysis of transcriptomic and genomic data of APOE44 vs. APOE33 allele carriers affected by Alzheimer’s disease. Network analysis combined with a novel technique for biomarker computation allowed the identification of an alteration in aging-associated processes such as inflammation, oxidative stress and metabolic pathways, indicating that APOE4 possibly accelerates pathological processes physiologically induced by aging. Subsequent integration with genomic data indicates that the Notch pathway could be the nodal molecular mechanism altered in APOE44 allele carriers with Alzheimer’s disease. Interestingly, PSEN1 and APP, genes whose mutation are known to be linked to early onset Alzheimer’s disease, are closely linked to this pathway. In conclusion, APOE4 role on inflammation and oxidation through the Notch signaling pathway could be crucial in elucidating the risk factors of Alzheimer’s disease.

Sequential forward-inverse design for genetic network modeling

Abstract This paper proposes methods for forward and inverse system modeling using Bayesian and least squares regression. These methods are based on both space-filling design criteria for multiple response problems and linear optimality criteria focusing on D-optimality. Modeling with and without the constant term is considered motivated by the case study application of genetic network modeling. We propose extended one-factor-at-a-time experimentation followed by augmentation of next stage design which offers biologists simplicity. Results are illustrated both numerical examples, a test problem from the literature, and a case study motivated by an real world biological research related to genetic network modeling.

Rank-based miRNA signatures for blood-based diagnosis of tuberculosis

We describe a new signature definition and analysis method to be used as biomarker for blood-based diagnosis of tuberculosis. Our new approach is based on the construction of a reference map of transcriptional signatures of both healthy and affected individuals using circulating miRNA from a large number of subjects. Once such a map is available, the diagnosis for a new patient can be performed by observing the relative position on the map of his/her transcriptional signature. To demonstrate its efficacy for this specific application we report the results of the application of our method to published data sets of circulating miRNA. Two crucial features make this method an ideal candidate for large scale applications such as a mass screening tool, or for point-of-care diagnostics. Specifically, our method is minimally invasive because it works well with profiles of circulating miRNA. More importantly, it is robust with respect to lab-to-lab protocol variability, measurement errors and batch effects because it requires that only the relative ranking of miRNA species in a profile be accurate, not their absolute values.