Machine learning, artificial intelligence, and other modern statistical methods are providing new opportunities to operationalise previously untapped and rapidly growing sources of data for patient benefit. Despite much promising research currently being undertaken, particularly in imaging, the literature as a whole lacks transparency, clear reporting to facilitate replicability, exploration for potential ethical concerns, and clear demonstrations of effectiveness. Among the many reasons why these problems exist, one of the most important (for which we provide a preliminary solution here) is the current lack of best practice guidance specific to machine learning and artificial intelligence. However, we believe that interdisciplinary groups pursuing research and impact projects involving machine learning and artificial intelligence for health would benefit from explicitly addressing a series of questions concerning transparency, reproducibility, ethics, and effectiveness (TREE). The 20 critical questions proposed here provide a framework for research groups to inform the design, conduct, and reporting; for editors and peer reviewers to evaluate contributions to the literature; and for patients, clinicians and policy makers to critically appraise where new findings may deliver patient benefit.

Motivation: The study of interactomes, or networks of protein-protein interactions, is increasingly providing valuable information on biological systems. Here we report a study of cancer proteins in an extensive human protein-protein interaction network constructed by computational methods. Results: We show that human proteins translated from known cancer genes exhibit a network topology that is different from that of proteins not documented as being mutated in cancer. In particular, cancer proteins show an increase in the number of proteins they interact with. They also appear to participate in central hubs rather than peripheral ones, mirroring their greater centrality and participation in networks that form the backbone of the proteome. Moreover, we show that cancer proteins contain a high ratio of highly promiscuous structural domains, i.e., domains with a high propensity for mediating protein interactions. These observations indicate an underlying evolutionary distinction between the two groups of proteins, reflecting the central roles of proteins, whose mutations lead to cancer. Contact:paul.bates@cancer.org.uk Supplementary information: The interactome data are available though the PIP (Potential Interactions of Proteins) web server at Author Webpage. Further additional material is available at Author Webpage

Background: Protein-protein interactions have traditionally been studied on a small scale, using classical biochemical methods to investigate the proteins of interest. More recently large-scale methods, such as two-hybrid screens, have been utilised to survey extensive portions of genomes. Current high-throughput approaches have a relatively high rate of errors, whereas in-depth biochemical studies are too expensive and time-consuming to be practical for extensive studies. As a result, there are gaps in our knowledge of many key biological networks, for which computational approaches are particularly suitable. Results: We constructed networks, or 'interactomes', of putative protein-protein interactions in the rat proteome – the rat being an organism extensively used for cancer studies. This was achieved by integrating experimental protein-protein interaction data from many species and translating this data into the reference frame of the rat. The putative rat protein interactions were given confidence scores based on their homology to proteins that have been experimentally observed to interact. The confidence score was furthermore weighted according to the extent of the experimental evidence, giving a higher weight to more frequently observed interactions. The scoring function was subsequently validated and networks constructed around key proteins, identified as being highly up- or down-regulated in rat cell lines of high metastatic potential. Using clustering methods on the networks, we have identified key protein communities involved in cancer metastasis. Conclusion: The protein network generation and subsequent network analysis used here, were shown to be useful for highlighting key proteins involved in metastasis. This approach, in conjunction with microarray expression data, can be extended to other species, thereby suggesting possible pathways around proteins of interest.

Background Reimbursement decisions are conventionally based on evidence from randomised controlled trials (RCTs), which often have high internal validity but low external validity. Real-world data (RWD) may provide complimentary evidence for relative effectiveness assessments (REAs) and cost-effectiveness assessments (CEAs). This study examines whether RWD is incorporated in health technology assessment (HTA) of melanoma drugs by European HTA agencies, as well as differences in RWD use between agencies and across time. Methods HTA reports published between 1 January 2011 and 31 December 2016 were retrieved from websites of agencies representing five jurisdictions: England [National Institute for Health and Care Excellence (NICE)], Scotland [Scottish Medicines Consortium (SMC)], France [Haute Autorité de santé (HAS)], Germany [Institute for Quality and Efficacy in Healthcare (IQWiG)] and The Netherlands [Zorginstituut Nederland (ZIN)]. A standardized data extraction form was used to extract information on RWD inclusion for both REAs and CEAs. Results Overall, 52 reports were retrieved, all of which contained REAs; CEAs were present in 25 of the reports. RWD was included in 28 of the 52 REAs (54%), mainly to estimate melanoma prevalence, and in 22 of the 25 (88%) CEAs, mainly to extrapolate long-term effectiveness and/or identify drug-related costs. Differences emerged between agencies regarding RWD use in REAs; the ZIN and IQWiG cited RWD for evidence on prevalence, whereas the NICE, SMC and HAS additionally cited RWD use for drug effectiveness. No visible trend for RWD use in REAs and CEAs over time was observed. Conclusion In general, RWD inclusion was higher in CEAs than REAs, and was mostly used to estimate melanoma prevalence in REAs or to predict long-term effectiveness in CEAs. Differences emerged between agencies’ use of RWD; however, no visible trends for RWD use over time were observed.

The prevalence of type 2 diabetes (T2DM) is increasing, but increasing longevity among persons with diagnosed diabetes may be is associated with more extensive and diverse types of morbidity. The extent and breadth of morbidity and how this varies across sub-groups is unclear and could have important clinical and public health implications. We aimed to estimate comorbidity profiles in people with T2DM and variations across sub-groups and over time. We identified approximately 224,000 people with T2DM in the Discover-NOW dataset, a real-world primary care database from 2000 to 2020 covering 2.5 million people across North-West London, England, linked to hospital records. We generated a mixed prevalence and incidence study population through repeated annual cross sections, and included a broad set of 35 comorbidities covering traditional T2DM conditions, emerging T2DM conditions and other common conditions. We estimated annual age-standardised prevalence of comorbidities, over the course of the disease in people with T2DM and several sub-groups. Multimorbidity (two or more chronic conditions) is common in people with T2DM and increasing, but the comorbidity profiles of people with T2DM vary substantially. Nearly 30% of T2DM patients had three or more comorbidities at diagnosis, increasing to 60% of patients ten years later. Two of the five commonest comorbidities at diagnosis were traditional T2DM conditions (hypertension (37%) and ischaemic heart disease (10%)) the other three were not (depression (15%), back pain (25%) and osteoarthritis (11%)). The prevalence of each increased during the course of the disease, with more than one in three patients having back pain and one in four having depression ten years post diagnosis. People with five or more comorbidities at diagnosis had higher prevalence of each of the 35 comorbidities. Hypertension (73%) was the commonest comorbidity at diagnosis in this group; followed by back pain (69%), depression (67%), asthma (45%) and osteoarthritis (36%). People with obesity at diagnosis had substantially different comorbidity profiles to those without, and the five commonest comorbidities were 50% more common in this group. Preventative and clinical interventions alongside care pathways for people with T2DM should transition to reflect the diverse set of causes driving persistent morbidity. This would benefit both patients and healthcare systems alike. The study was funded by the National Institute for Health and Care Excellence (NICE).

To reduce harm to the environment resulting from the production, use, and disposal of health technologies, there are different options for how health technology assessment (HTA) agencies can consider environmental information. We identified four approaches that HTA agencies can use to take environmental information into account in healthcare decision making and the challenges associated with each approach. Republishing data that is in the public domain or has been submitted to an HTA agency we term the “information conduit” approach. Analyzing and presenting environmental data separately from established health economic analyses is described as “parallel evaluation.” Integrating environmental impact into HTAs by identifying or creating new methods that allow clinical, financial, and environmental information to be combined in a single quantitative analysis is “integrated evaluation.” Finally, evidence synthesis and analysis of health technologies that are not expected to improve health-related outcomes but claim to have relative environmental benefits are termed “environment-focused evaluation.”