Dieter Galea

Mitochondria mediate septin cage assembly to promote autophagy of Shigella

Septins, cytoskeletal proteins with well‐characterised roles in cytokinesis, form cage‐like structures around cytosolic Shigella flexneri and promote their targeting to autophagosomes. However, the processes underlying septin cage assembly, and whether they influence S. flexneri proliferation, remain to be established. Using single‐cell analysis, we show that the septin cages inhibit S. flexneri proliferation. To study mechanisms of septin cage assembly, we used proteomics and found mitochondrial proteins associate with septins in S. flexneri‐infected cells. Strikingly, mitochondria associated with S. flexneri promote septin assembly into cages that entrap bacteria for autophagy. We demonstrate that the cytosolic GTPase dynamin‐related protein 1 (Drp1) interacts with septins to enhance mitochondrial fission. To avoid autophagy, actin‐polymerising Shigella fragment mitochondria to escape from septin caging. Our results demonstrate a role for mitochondria in anti‐Shigella autophagy and uncover a fundamental link between septin assembly and mitochondria.

Translational utility of a hierarchical classification strategy in biomolecular data analytics

Hierarchical classification (HC) stratifies and classifies data from broad classes into more specific classes. Unlike commonly used data classification strategies, this enables the probabilistic prediction of unknown classes at different levels, minimizing the burden of incomplete databases. Despite these advantages, its translational application in biomedical sciences has been limited. We describe and demonstrate the implementation of a HC approach for “omics-driven” classification of 15 bacterial species at various taxonomic levels achieving 90–100% accuracy, and 9 cancer types into morphological types and 35 subtypes with 99% and 76% accuracy, respectively. Unknown bacterial species were probabilistically assigned with 100% accuracy to their respective genus or family using mass spectra (n = 284). Cancer types were predicted by mRNA data (n = 1960) for most subtypes with 95–100% accuracy. This has high relevance in clinical practice where complete datasets are difficult to compile with the continuous evolution of diseases and emergence of new strains, yet prediction of unknown classes, such as bacterial species, at upper hierarchy levels may be sufficient to initiate antimicrobial therapy. The algorithms presented here can be directly translated into clinical-use with any quantitative data, and have broad application potential, from unlabeled sample identification, to hierarchical feature selection, and discovery of new taxonomic variants.

Exosomal microRNAs derived from colorectal cancer-associated fibroblasts: role in driving cancer progression

Colorectal cancer is a global disease with increasing incidence. Mortality is largely attributed to metastatic spread and therefore, a mechanistic dissection of the signals which influence tumor progression is needed. Cancer stroma plays a critical role in tumor proliferation, invasion and chemoresistance. Here, we sought to identify and characterize exosomal microRNAs as mediators of stromal-tumor signaling. In vitro, we demonstrated that fibroblast exosomes are transferred to colorectal cancer cells, with a resultant increase in cellular microRNA levels, impacting proliferation and chemoresistance. To probe this further, exosomal microRNAs were profiled from paired patient-derived normal and cancer-associated fibroblasts, from an ongoing prospective biomarker study. An exosomal cancer-associated fibroblast signature consisting of microRNAs 329, 181a, 199b, 382, 215 and 21 was identified. Of these, miR-21 had highest abundance and was enriched in exosomes. Orthotopic xenografts established with miR-21-overexpressing fibroblasts and CRC cells led to increased liver metastases compared to those established with control fibroblasts. Our data provide a novel stromal exosome signature in colorectal cancer, which has potential for biomarker validation. Furthermore, we confirmed the importance of stromal miR-21 in colorectal cancer progression using an orthotopic model, and propose that exosomes are a vehicle for miR-21 transfer between stromal fibroblasts and cancer cells.

BASIS : High-performance bioinformatics platform for processing of large-scale mass spectrometry imaging data in chemically augmented histology

Mass Spectrometry Imaging (MSI) holds significant promise in augmenting digital histopathologic analysis by generating highly robust big data about the metabolic, lipidomic and proteomic molecular content of the samples. In the process, a vast quantity of unrefined data, that can amount to several hundred gigabytes per tissue section, is produced. Managing, analysing and interpreting this data is a significant challenge and represents a major barrier to the translational application of MSI. Existing data analysis solutions for MSI rely on a set of heterogeneous bioinformatics packages that are not scalable for the reproducible processing of large-scale (hundreds to thousands) biological sample sets. Here, we present a computational platform (pyBASIS) capable of optimized and scalable processing of MSI data for improved information recovery and comparative analysis across tissue specimens using machine learning and related pattern recognition approaches. The proposed solution also provides a means of seamlessly integrating experimental laboratory data with downstream bioinformatics interpretation/analyses, resulting in a truly integrated system for translational MSI.

Exploiting and assessing multi-source data for supervised biomedical named entity recognition

Motivation: Recognition of biomedical entities from scientific text is a critical component of natural language processing and automated information extraction platforms. Modern named entity recognition approaches rely heavily on supervised machine learning techniques, which are critically dependent on annotated training corpora. These approaches have been shown to perform well when trained and tested on the same source. However, in such scenario, the performance and evaluation of these models may be optimistic, as such models may not necessarily generalize to independent corpora, resulting in potential non‐optimal entity recognition for large‐scale tagging of widely diverse articles in databases such as PubMed. Results: Here we aggregated published corpora for the recognition of biomolecular entities (such as genes, RNA, proteins, variants, drugs and metabolites), identified entity class overlap and performed leave‐corpus‐out cross validation strategy to test the efficiency of existing models. We demonstrate that accuracies of models trained on individual corpora decrease substantially for recognition of the same biomolecular entity classes in independent corpora. This behavior is possibly due to limited generalizability of entity‐class‐related features captured by individual corpora (model ‘overtraining’) which we investigated further at the orthographic level, as well as potential annotation standard differences. We show that the combined use of multi‐source training corpora results in overall more generalizable models for named entity recognition, while achieving comparable individual performance. By performing learning‐curve‐based power analysis we further identified that performance is often not limited by the quantity of the annotated data. Availability and implementation: Compiled primary and secondary sources of the aggregated corpora are available on: https://github.com/dterg/biomedical_corpora/wiki and https://bitbucket.org/iAnalytica/bioner. Supplementary information: Supplementary data are available at Bioinformatics online.

Statistical Tools for Molecular Covariance Spectroscopy

One major application of modern spectroscopic and spectrometric techniques is to measure hundreds to thousands of molecules in biological specimens as part of a process of metabolic phenotyping. Statistical spectroscopy covers a range of techniques used for the recovery of correlated intensity patterns within and between molecules. This plays an essential role in the annotation of molecular features of potential biological or diagnostic significance. The article introduces a variety of univariate and multivariate statistical tools for molecular covariance spectroscopy.

Abstract 839: Network-driven analytics of published tissue-based biomarkers to predict response to neoadjuvant therapy in rectal cancer

Introduction Neoadjuvant chemoradiotherapy (nCRT) plays a central role in the management of locally advanced rectal cancer. For many, nCRT leads to clinically meaningful tumour regression. However, up to 20% exhibit little to no response and, in this group, nCRT results in unnecessary delays to definitive treatment. There is a critical need for development of robust molecular methods to predict response to nCRT, to allow for more precise treatment stratification. Although numerous molecular pathways and biomarkers have been implicated in radiosensitivity, the lack of a unifying interpretation of these findings has restricted translational deployment. The aim of the current study was to develop a ‘knowledge network’ with which to visualise and interpret published, quantitative, biomarker data relating to radiosensitivity in rectal cancer, beyond the conventional format of a systematic review. Methods Existing data on predictive biomarkers were retrieved by way of a systematic review of electronic bibliographic databases. Biomarkers were classified according to biological function and built into a hierarchical Gene Ontology tree. Significance was binarized based on p-values or multivariate statistics. An interactive, direct acyclic graph was developed using the Dagre-D3 JavaScript library. Nodes were sized by number of studied biomarkers and color-coded according to their significance scores. The scores reflect the ratio of significant versus non-significant evidence across studied biomarkers. A negative score range indicates more non-significant biomarker findings for that ontological term (node). Weightings were applied to reflect those biomarkers confirmed as significant across two or more studies. p-values of 0.05 or less (adjusted for multiple comparative analysis where appropriate) were considered to be statistically significant. Results 72 individual biomarkers were identified through review. On highest order classification, the domains of response to stress and factors inhibiting apoptosis were found to be most significant (aggregate significance scores across identified biomarkers, 0.75 and 0.714 respectively). A predictive power was not reached for the majority of prognostic biomarkers; rather, the levels of their statistical significance were assessed. Conclusions Building a knowledge-based network analysis of published data identifies promising areas for further research into cellular mechanisms, which may aid in biomarker discovery. Regarding significant node clusters within a network of published data on predictive biomarkers, modifications in cellular metabolic responses to the insult posed by nCRT appear to hold promise in developing a panel of biomarkers with a predictive capacity for response. Network-based analytics takes into account the complex nature of response to therapy, and is a novel way of presenting results obtained from a systematic review. Citation Format: Liam R. Poynter, Kirill Veselkov, Dieter Galea, James Kinross, Alexander Mirnezami, Jeremy Nicholson, Zoltan Takats, Reza Mirnezami, Ara Darzi. Network-driven analytics of published tissue-based biomarkers to predict response to neoadjuvant therapy in rectal cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2017; 2017 Apr 1-5; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2017;77(13 Suppl):Abstract nr 839. doi:10.1158/1538-7445.AM2017-839