Interpretable Link Prediction in AI-Driven Cancer Research: Uncovering Co-Authorship Patterns

Interpretable Link Prediction in AI-Driven Cancer Research: Uncovering Co-Authorship Patterns Shahab Mosallaie1, Andrea Schiffauerova1 and Ashkan Ebadi1,2,*

1 CIISE, Concordia University, Canada 2 Digital Technologies, National Research Council Canada, Canada.

• Email: ashkan.ebadi@nrc-cnrc.gc.ca Abstract Artificial intelligence (AI) is transforming cancer diagnosis and treatment. The intricate nature of this disease necessitates the collaboration of diverse stakeholders with varied expertise to ensure the effectiveness of cancer research. Despite its importance, forming effective interdisciplinary research teams remains challenging. Understanding and predicting collaboration patterns can help researchers, organizations, and policymakers optimize resources and foster impactful research. We examined co- authorship networks as a proxy for collaboration within AI-driven cancer research. Using 7,738 publications (2000-2017) from Scopus, we constructed 36 overlapping co-authorship networks representing new, persistent, and discontinued collaborations. We engineered both attribute-based and structure-based features and built four machine learning classifiers. Model interpretability was performed using Shapley Additive Explanations (SHAP). Random forest achieved the highest recall for all three types of examined collaborations. The discipline similarity score emerged as a crucial factor, positively affecting new and persistent patterns while negatively impacting discontinued collaborations. Additionally, high productivity and seniority were positively associated with discontinued links. Our findings can guide the formation of effective research teams, enhance interdisciplinary cooperation, and inform strategic policy decisions. Keywords Collaborative patterns, Link prediction, Machine learning, Interpretability, Cancer

1. Introduction Cancer remains a leading cause of death globally, accounting for nearly 10 million fatalities in 2020 (Ferlay et al. 2021), representing approximately one in every six deaths recorded worldwide (World Health Organization 2022). In Canada, cancer continues to hold its position as the primary cause of death, with approximately 40% of Canadians expected to receive a cancer diagnosis in their lifetime, and about 25% likely to succumb to the disease (Government of Canada 2021). Despite its lethal nature, early detection and effective treatment can lead to the successful cure of many patients and cancer types (World Health Organization 2022). In fact, timely diagnosis and intervention can arrest cancer progression and enhance prognosis (Crosby et al. 2022). This highlights the urgent necessity for effective cancer diagnostic strategies and procedures to save countless lives from this life-threatening disease (Okoli et al. 2021). The rapid evolution of advanced cancer treatment technologies further underscores the importance of interdisciplinary collaboration. For example, nanoparticle-based drug delivery systems have emerged as a promising approach to improve therapeutic precision and reduce systemic toxicity in oncology. These systems can be engineered to target specific tumour sites, enabling controlled release of therapeutic agents and potentially improving patient outcomes (Cheng et al. 2021). The development of such technologies requires close collaboration among oncologists, materials scientists, biomedical engineers, and data scientists, particularly in the design, testing, and optimization phases.

2

Enhanced health outcomes can be achieved through collaboration among healthcare professionals (Ebadi et al. 2017), facilitated by knowledge sharing and improved decision-making. Cancer care, being a complex and fragmented process (Ullgren 2021), requires healthcare providers and patients to work together as a cohesive team to deliver high-quality care (Amafah et al. 2023). Furthermore, effective cancer care, encompassing both diagnosis and treatment, requires collaboration among researchers and professionals from diverse disciplines, each contributing their unique expertise (Knoop, Wujcik, and Wujcik 2017). Many AI-driven healthcare research projects are inherently international in scope, bringing together multi-country teams that integrate diverse expertise, resources, and perspectives to address complex scientific and clinical challenges. Considerable effort has been dedicated to understanding the dynamics of research collaboration and pinpointing the factors that encourage it. Several studies credited innovation and scientific discoveries to the collaborative efforts of individual researchers (e.g., Beck et al. 2022). Studies also highlighted the positive impact of collaboration on research productivity (e.g., Ebadi and Schiffauerova 2016). Co-authorship is frequently used as a proxy for scientific collaboration (Ebadi and Schiffauerova 2015; Essers, Grigoli, and Pugacheva 2022), a trend further facilitated by t