Unsupervised construction of computational graphs for gene expression data with explicit structural inductive biases.

MOTIVATION: Gene expression data are commonly used at the intersection of cancer research and machine learning for better understanding of the molecular status of tumour tissue. Deep learning predictive models have been employed for gene expression data due to their ability to scale and remove the need for manual feature engineering. However, gene expression data are often very high dimensional, noisy and presented with a low number of samples. This poses significant problems for learning algorithms: models often overfit, learn noise and struggle to capture biologically relevant information. In this article, we utilize external biological knowledge embedded within structures of gene interaction graphs such as protein-protein interaction (PPI) networks to guide the construction of predictive models. RESULTS: We present Gene Interaction Network Constrained Construction (GINCCo), an unsupervised method for automated construction of computational graph models for gene expression data that are structurally constrained by prior knowledge of gene interaction networks. We employ this methodology in a case study on incorporating a PPI network in cancer phenotype prediction tasks. Our computational graphs are structurally constructed using topological clustering algorithms on the PPI networks which incorporate inductive biases stemming from network biology research on protein complex discovery. Each of the entities in the GINCCo computational graph represents biological entities such as genes, candidate protein complexes and phenotypes instead of arbitrary hidden nodes of a neural network. This provides a biologically relevant mechanism for model regularization yielding strong predictive performance while drastically reducing the number of model parameters and enabling guided post-hoc enrichment analyses of influential gene sets with respect to target phenotypes. Our experiments analysing a variety of cancer phenotypes show that GINCCo often outperforms support vector machine, Fully Connected Multi-layer Perceptrons (MLP) and Randomly Connected MLPs despite greatly reduced model complexity. AVAILABILITY AND IMPLEMENTATION: https://github.com/paulmorio/gincco contains the source code for our approach. We also release a library with algorithms for protein complex discovery within PPI networks at https://github.com/paulmorio/protclus. This repository contains implementations of the clustering algorithms used in this article. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Human inference beyond syllogisms: an approach using external graphical representations.

Research in psychology about reasoning has often been restricted to relatively inexpressive statements involving quantifiers (e.g. syllogisms). This is limited to situations that typically do not arise in practical settings, like ontology engineering. In order to provide an analysis of inference, we focus on reasoning tasks presented in external graphic representations where statements correspond to those involving multiple quantifiers and unary and binary relations. Our experiment measured participants' performance when reasoning with two notations. The first notation used topological constraints to convey information via node-link diagrams (i.e. graphs). The second used topological and spatial constraints to convey information (Euler diagrams with additional graph-like syntax). We found that topo-spatial representations were more effective for inferences than topological representations alone. Reasoning with statements involving multiple quantifiers was harder than reasoning with single quantifiers in topological representations, but not in topo-spatial representations. These findings are compared to those in sentential reasoning tasks.

Work-related Hazards Among Pathologists and Residents of Pathology:Results of a Cross-sectional Study in Iran.

BACKGROUND & OBJECTIVE: Pathologists as medical professionals involved in the diagnosis and planning of therapies in many diseases are exposed to occupational hazards in workplaces. Hence, we aimed to determine the occupational health problems among Iranian pathologists in this cross-sectional study. METHODS: This cross-sectional study was conducted among the Iranian pathologists. The data required for this study was collected through a self-reported questionnaire containing 48 questions about major occupational health problems, including musculoskeletal problems, visual disorders, workplace characteristics, health behavior, and other medical conditions. RESULTS: Among the study participants (N=350), 87.4% presented with musculoskeletal disorders in the past year, with the neck as the most common location of pain (71%). Musculoskeletal pain was significantly higher in those working with the computer for more than 5 hours per day (P=0.007). Furthermore, 273 (78%) participants reported visual refractive errors, and myopia was the most common error (53%). Acute injuries were reported in 263 (75%) participants, and the cutting injury had the highest frequency (56.6%). Depression was reported in 54 (15.4%) of the participants, followed by burnout (10.3%) and hypertension (4%). Intolerance reactions to formalin were reported by 222 (63.6%) and were significantly more frequent among the residents (P<0.001). The residents were more prone to musculoskeletal pain (P=0.002) and injury (P=0.026). CONCLUSION: We observed a noticeable prevalence of health risks, including musculoskeletal problems, visual disturbances, injuries, and ergonomic problems among the Iranian pathologists. Solving these problems demands thorough prevention and personal protection, as well as educational programs with more attention toward optimization of ergonomics in the workplace and awareness about chemical and biological hazards.

Expressions of TWIST1 and CD105 markers in colorectal cancer patients and their association with metastatic potential and prognosis.

BACKGROUND: TWIST1 and CD105, which contribute to tumor malignancy, are overexpressed in cancers. Accordingly, TWIST1 enhances epithelial-to-mesenchymal transition (EMT) and promotes the formation of cancer stem cells (CSCs). Also, CD105 is a neoangiogenesis marker in endothelial cells, which is introduced as a CSC marker in tumoral epithelial cells in several types of cancers. The present study was aimed to investigate expressions of TWIST1 and CD105 in colorectal cancer (CRC) patients. METHODS: Expressions of TWIST1 and CD105 in 250 CRC tissue samples were evaluated using immunohistochemistry on tissue microarrays (TMAs). In this regard, TWIST1 expression was investigated in the subcellular locations (cytoplasm and nucleus), while CD105 was mapped in endothelial cells and cytoplasmic tumor cells of CRC tissues. The association between the expression of these markers and clinicopathological parameters, as well as survival outcomes were analyzed. RESULTS: Results indicate a statistically significant association between higher nuclear expression levels of TWIST1 and distant metastases in CRC (P = 0.040) patients. In addition, it was shown that the increased nuclear expression of TWIST1 had a poor prognostic value for disease-specific survival (DSS) and progression-free survival (PFS) (P = 0.042, P = 0.043, respectively) in patients with CRC. Moreover, analysis of CD105 expression level has revealed that there is a statistically significant association between the increased expression of CD105 in tumoral epithelial cells and more advanced TNM stage (P = 0.050). CONCLUSIONS: Our results demonstrate that nuclear TWIST1 and cytoplasmic CD105 expressions in tumor cells had associations with more aggressive tumor behavior and more advanced diseases in CRC cases.

High expression of DNA damage-inducible transcript 4 (DDIT4) is associated with advanced pathological features in the patients with colorectal cancer.

DNA damage-inducible transcript 4 (DDIT4) is induced in various cellular stress conditions. This study was conducted to investigate expression and prognostic significance of DDIT4 protein as a biomarker in the patients with colorectal cancer (CRC). PPI network and KEGG pathway analysis were applied to identify hub genes among obtained differentially expressed genes in CRC tissues from three GEO Series. In clinical, expression of DDIT4 as one of hub genes in three subcellular locations was evaluated in 198 CRC tissues using immunohistochemistry method on tissue microarrays. The association between DDIT4 expression and clinicopathological features as well as survival outcomes were analyzed. Results of bioinformatics analysis indicated 14 hub genes enriched in significant pathways according to KEGG pathways analysis among which DDIT4 was selected to evaluate CRC tissues. Overexpression of nuclear DDIT4 protein was found in CRC tissues compared to adjacent normal tissues (P = 0.003). Furthermore, higher nuclear expression of DDIT4 was found to be significantly associated with the reduced tumor differentiation and advanced TNM stages (all, P = 0.009). No significant association was observed between survival outcomes and nuclear expression of DDIT4 in CRC cases. Our findings indicated higher nuclear expression of DDIT4 was significantly associated with more aggressive tumor behavior and more advanced stage of disease in the patients with CRC.

Low expression of Talin1 is associated with advanced pathological features in colorectal cancer patients.

To explore the proper prognostic markers for the likelihood of metastasis in CRC patients. Seventy-seven fresh CRC samples were collected to evaluate the mRNA level of the selected marker using Real-time PCR. Moreover, 648 formalin-fixed paraffin-embedded CRC tissues were gathered to evaluate protein expression by immunohistochemistry (IHC) on tissue microarrays. The results of Real-Time PCR showed that low expression of Talin1 was significantly associated with advanced TNM stage (p = 0.034) as well as gender (p = 0.029) in mRNA levels. Similarly, IHC results indicated that a low level of cytoplasmic expression of Talin1 was significantly associated with advanced TNM stage (p = 0.028) as well as gender (p = 0.009) in CRC patients. Moreover, decreased expression of cytoplasmic Talin1 protein was found to be a significant predictor of worse disease-specific survival (DSS) (p = 0.011) in the univariate analysis. In addition, a significant difference was achieved (p = 0.039) in 5-year survival rates of DSS: 65% for low, 72% for moderate, and 88% for high Talin1 protein expression. Observations showed that lower expression of Talin1 at both the gene and protein level may drive the disparity of CRC patients' outcomes via worse DSS and provide new insights into the development of progression indicators because of its correlation with increased tumor aggressiveness.

Variational Autoencoders for Cancer Data Integration: Design Principles and Computational Practice.

International initiatives such as the Molecular Taxonomy of Breast Cancer International Consortium are collecting multiple data sets at different genome-scales with the aim to identify novel cancer bio-markers and predict patient survival. To analyze such data, several machine learning, bioinformatics, and statistical methods have been applied, among them neural networks such as autoencoders. Although these models provide a good statistical learning framework to analyze multi-omic and/or clinical data, there is a distinct lack of work on how to integrate diverse patient data and identify the optimal design best suited to the available data.In this paper, we investigate several autoencoder architectures that integrate a variety of cancer patient data types (e.g., multi-omics and clinical data). We perform extensive analyses of these approaches and provide a clear methodological and computational framework for designing systems that enable clinicians to investigate cancer traits and translate the results into clinical applications. We demonstrate how these networks can be designed, built, and, in particular, applied to tasks of integrative analyses of heterogeneous breast cancer data. The results show that these approaches yield relevant data representations that, in turn, lead to accurate and stable diagnosis.