Explainable discovery of disease biomarkers: The case of ovarian cancer to illustrate the best practice in machine learning and Shapley analysis.

OBJECTIVE: Ovarian cancer is a significant health issue with lasting impacts on the community. Despite recent advances in surgical, chemotherapeutic and radiotherapeutic interventions, they have had only marginal impacts due to an inability to identify biomarkers at an early stage. Biomarker discovery is challenging, yet essential for improving drug discovery and clinical care. Machine learning (ML) techniques are invaluable for recognising complex patterns in biomarkers compared to conventional methods, yet they can lack physical insights into diagnosis. eXplainable Artificial Intelligence (XAI) is capable of providing deeper insights into the decision-making of complex ML algorithms increasing their applicability. We aim to introduce best practice for combining ML and XAI techniques for biomarker validation tasks. METHODS: We focused on classification tasks and a game theoretic approach based on Shapley values to build and evaluate models and visualise results. We described the workflow and apply the pipeline in a case study using the CDAS PLCO Ovarian Biomarkers dataset to demonstrate the potential for accuracy and utility. RESULTS: The case study results demonstrate the efficacy of the ML pipeline, its consistency, and advantages compared to conventional statistical approaches. CONCLUSION: The resulting guidelines provide a general framework for practical application of XAI in medical research that can inform clinicians and validate and explain cancer biomarkers.

Chinese newborn screening for the incidence of G6PD deficiency and variant of G6PD gene from 2013 to 2017.

Glucose-6-phosphate dehydrogenase (G6PD) deficiency is one of the most common X-linked enzymopathies caused by G6PD gene variant. We aimed to provide the characteristics of G6PD deficiency and G6PD gene variant distribution in a large Chinese newborn screening population. We investigated the prevalence of G6PD in China from 2013 to 2017. Then, we examined G6PD activity and G6PD gene in representative Chinese birth cohort to explore the distribution of G6PD gene variant in 2016. We then performed multicolor melting curve analysis to classify G6PD gene variants in 10,357 neonates with activity-confirmed G6PD deficiency, and DNA Sanger sequencing for G6PD coding exons if hot site variants were not found. The screened population, organizations, and provinces of G6PD deficiency were increased from 2013 to 2017 in China. The top five frequency of G6PD gene variants were c.1376G>T, c.1388G>A, c.95A>G, c.1024C>T, and c.871G>A and varied in different provinces, with regional and ethnic features, and four pathogenic variant sites (c.152C>T, c.290A>T, c.697G>C, and c.1285A>G) were first reported. G6PD deficiency mainly occurs in South China, and the frequency of G6PD gene variant varies in different regions and ethnicities.

Association of plasma retinol levels with incident cancer risk in Chinese hypertensive adults: a nested case-control study.

We aimed to investigate the association between plasma retinol and incident cancer among Chinese hypertensive adults. We conducted a nested case-control study, including 231 patients with incident cancer and 231 matched controls during a median 4·5-year follow-up of the China Stroke Primary Prevention Trial. There was a significant, inverse association between retinol levels and digestive system cancer (per 10 µg/dl increases: OR 0·79; 95 % CI 0·69, 0·91). When compared with participants in the first quartile of retinol (< 52·3 µg/dl), a significantly lower cancer risk was found in participants in quartile 2-4 ( ≥ 52·3 µg/dl: OR 0·31; 95 % CI 0·13, 0·71). However, there was a U-shaped association between retinol levels and non-digestive system cancers where the risk of cancers decreased (although not significantly) with each increment of plasma retinol (per 10 µg/dl increases: OR 0·89; 95 % CI 0·60, 1·31) in participants with retinol < 68·2 µg/dl, and then increased significantly with retinol (per 10 µg/dl increase: OR 1·65; 95 % CI 1·12, 2·44) in participants with retinol ≥ 68·2 µg/dl. In conclusion, there was a significant inverse dose-response association between plasma retinol and the risk of digestive system cancers. However, a U-shaped association was observed between plasma retinol and the risk of non-digestive cancers (with a turning point approximately 68·2 µg/dl).

Research on the impact of pilot free trade zones on urban green development: A case study based on the Yangtze River Economic Belt in China.

Green development is an important component of China's new development concept. Pilot Free Trade Zones (PFTZs), as "experimental fields" for promoting reform, deepening opening-up, and raising the level of an open economy, are important open areas for China to promote green development. However, existing related research is not extensive. This article takes PFTZs as quasi-natural experiments, with the Yangtze River Economic Belt (YREB) as the research area. Based on urban panel data from 2006 to 2020, using multi-period differences-in-differences and spatial differences-in-differences models, it explores the impact effects of PFTZs on urban green development and their potential mechanisms. The research findings indicate: (1) Overall, PFTZs significantly promote urban green development, with variations in impact effects due to different batches and locations of establishment. (2) Mechanism tests show that PFTZs mainly promote urban green development by stimulating technological innovation, industrial upgrading, and reducing government intervention. (3) From the perspective of spatial spillover effects, the establishment of PFTZs not only promotes the green development process in the host cities but also has a promoting effect on the green development of surrounding cities.

Superporous sponge prepared by secondary network compaction with enhanced permeability and mechanical properties for non-compressible hemostasis in pigs.

Developing superporous hemostatic sponges with simultaneously enhanced permeability and mechanical properties remains challenging but highly desirable to achieve rapid hemostasis for non-compressible hemorrhage. Typical approaches to improve the permeability of hemostatic sponges by increasing porosity sacrifice mechanical properties and yield limited pore interconnectivity, thereby undermining the hemostatic efficacy and subsequent tissue regeneration. Herein, we propose a temperature-assisted secondary network compaction strategy following the phase separation-induced primary compaction to fabricate the superporous chitosan sponge with highly-interconnected porous structure, enhanced blood absorption rate and capacity, and fatigue resistance. The superporous chitosan sponge exhibits rapid shape recovery after absorbing blood and maintains sufficient pressure on wounds to build a robust physical barrier to greatly improve hemostatic efficiency. Furthermore, the superporous chitosan sponge outperforms commercial gauze, gelatin sponges, and chitosan powder by enhancing hemostatic efficiency, cell infiltration, vascular regeneration, and in-situ tissue regeneration in non-compressible organ injury models, respectively. We believe the proposed secondary network compaction strategy provides a simple yet effective method to fabricate superporous hemostatic sponges for diverse clinical applications.

Single nucleotide polymorphism and haplotype association of the interleukin-8 gene with nasopharyngeal carcinoma.

The cytokine interleukin-8 (IL-8) may play a role in the pathogenesis of nasopharyngeal carcinoma (NPC) through the modulation of tumor immune response or enhanced angiogenesis. Polymorphism of IL-8 gene, which may affect the production level of cytokine, has been inversely associated with a number of cancers. To test this hypothesis, we investigated the relationship of IL-8 gene polymorphisms and NPC in a Chinese population. We analyzed single nucleotide polymorphisms (SNPs) of IL-8 gene -845 T/C, -738 T/A, -353 A/T, -251 A/T and +678 T/C in 280 patients with NPC and 290 age and sex matched controls, using polymerase chain reaction-restriction fragment length polymorphism (PCR-RFLP) and polymerase chain reaction-sequence specific primers method (PCR-SSP). There were significant differences in the genotype and allele distribution of -251 A/T polymorphism of the IL-8 gene among cases and controls. The -251 AA and AT genotypes were associated with a significantly increased risk of NPC as compared with the -251 TT genotypes (OR=1.820, 95% CI, 1.120-2.959, P=0.015 and OR=1.590, 95% CI, 1.104-2.290, P=0.013, respectively). Haplotype analysis revealed that the homozygosity of the AAT haplotype (defined by SNPs at positions -353, -251 and +678) of IL-8 gene conveys the highest risk for NPC compared with the homozygosity for the TTC haplotype (OR=1.396; 95% CI, 1.064-1.831; P=0.016). The -251 A/T polymorphism of IL-8 and its haplotype are associated with NPC in a Chinese population. Our data suggests that IL-8 gene may play a role in the development of NPC.