[Rare disease in the age of artificial intelligence.] / Ammalarsi di una patologia rara in tempi di intelligenza artificiale.

INTRODUCTION: The text examines the impact of artificial intelligence (AI) in the context of rare diseases, exploring how patients turn to AI resources for health information, especially in situations where doctor-patient communication is limited. The article features the case of a doctor specializing in clinical psychology and psychotherapy, diagnosed with thymoma and Good's syndrome, who uses AI resources during his illness. METHODS: The capabilities of five chatbots based on Large Language Models (LLMs), such as GPT-3.5, GPT-4, Bing Chat, Google Bard, and Anthropic Claude are explored. The AIs were queried on various aspects of the disease, from pre-diagnosis and diagnosis to therapeutic, psychological, and caregiver management issues. The responses were evaluated by five experts based on criteria such as: accuracy, relevance, coherence, clarity, practical utility, ethical considerations, empathy, and capacity to respond to questions and concerns. RESULTS: The results indicate consistency in the evaluators' assessments, with generally high scores across all dimensions. Particularly, systems like Bard and GPT-4 received high ratings in terms of information accuracy and the ability to respond to questions and concerns. Bing and Claude were appreciated for their empathy and tone. Overall, the AI systems' responses were considered appropriate, respectful of ethics and privacy, and useful in the clinical context. DISCUSSION: The article emphasizes the importance of understanding the reliability and precision of responses provided by AI systems in the clinical field. Although these systems offer high-quality responses, there is significant variability in their performance. Healthcare professionals must be aware of these differences and use such tools cautiously. AI can provide support in some aspects of care but cannot replace genuine human empathy and understanding. Integrating AI into clinical practice presents potential but also challenges, particularly the possibility of providing incorrect information. CONCLUSIONS: The AI systems demonstrate the ability to provide useful advice on clinical and psychological issues, but their use requires caution. It is crucial to distinguish the benefits of AI for patients from the challenges it presents for healthcare professionals. As AI technology continues to evolve, it is essential that its integration into the clinical field is accompanied by continuous research and evaluations, to ensure safe and effective use in the healthcare sector.

A network approach for low dimensional signatures from high throughput data.

One of the main objectives of high-throughput genomics studies is to obtain a low-dimensional set of observables-a signature-for sample classification purposes (diagnosis, prognosis, stratification). Biological data, such as gene or protein expression, are commonly characterized by an up/down regulation behavior, for which discriminant-based methods could perform with high accuracy and easy interpretability. To obtain the most out of these methods features selection is even more critical, but it is known to be a NP-hard problem, and thus most feature selection approaches focuses on one feature at the time (k-best, Sequential Feature Selection, recursive feature elimination). We propose DNetPRO, Discriminant Analysis with Network PROcessing, a supervised network-based signature identification method. This method implements a network-based heuristic to generate one or more signatures out of the best performing feature pairs. The algorithm is easily scalable, allowing efficient computing for high number of observables ([Formula: see text]-[Formula: see text]). We show applications on real high-throughput genomic datasets in which our method outperforms existing results, or is compatible with them but with a smaller number of selected features. Moreover, the geometrical simplicity of the resulting class-separation surfaces allows a clearer interpretation of the obtained signatures in comparison to nonlinear classification models.

The genetic and genomic background of multiple myeloma patients achieving complete response after induction therapy with bortezomib, thalidomide and dexamethasone (VTD).

The prime focus of the current therapeutic strategy for Multiple Myeloma (MM) is to obtain an early and deep tumour burden reduction, up to the level of complete response (CR). To date, no description of the characteristics of the plasma cells (PC) prone to achieve CR has been reported. This study aimed at the molecular characterization of PC obtained at baseline from MM patients in CR after bortezomib-thalidomide-dexamethasone (VTD) first line therapy.One hundred and eighteen MM primary tumours obtained from homogeneously treated patients were profiled both for gene expression and for single nucleotide polymorphism genotype. Genomic results were used to obtain a predictor of sensitivity to VTD induction therapy, as well as to describe both the transcription and the genomic profile of PC derived from MM with subsequent optimal response to primary induction therapy.By analysing the gene profiles of CR patients, we identified a 5-gene signature predicting CR with an overall median accuracy of 75% (range: 72%-85%). In addition, we highlighted the differential expression of a series of genes, whose deregulation might explain patients' sensitivity to VTD therapy. We also showed that a small copy number loss, covering 606Kb on chromosome 1p22.1 was the most significantly associated with CR patients.