Artificial Intelligence and false data in biomedicine

Artificial intelligence (AI) is revolutionizing biomedicine, impacting key areas such as interpretation of medical images and the analysis of large research databases. AI is making important advances in fields such as radiology or blood analysis. histological sections (biopsies), fundus images or endoscopiestransforming the way medical professionals diagnose and treat diseases.

Current applications of AI in medicine

Today, AI-powered tools not only make it easier to capture medical dictations and transcribe patient interviews, but also enable automatic generation of notes in electronic medical records. These innovations in artificial intelligence applied to medicine They are no longer mere futuristic projections, but realities that benefit both doctors and patients in daily practice.

Big Data Analysis and Omic Sciences

Furthermore, AI is increasingly present in the analysis of massive databases (Big Data), ranging from clinical data to genomic results. Omic technologies, such as genome, transcriptome, proteome either metabolomegenerate large volumes of data that are difficult to analyze with traditional methods. This is where the AI and machine learning They play a crucial role, providing advanced tools for the analysis of this complex data.

Errors in data analysis and their consequences

A recent example shows the risks associated with errors in data analysis. In July 2024, Dr. Miguel Hernández Bronchud highlighted a case study in his post on LinkedIn where the prestigious magazine Nature had to retract an article due to an error in the classification of genomic sequences. The study authors confused millions of human DNA sequences with microbial sequences, resulting in incorrect conclusions about relationships between cancer DNA and microbiota.

Risks of noise and bias in Big Data analysis

This incident highlights the dangers of “noise” and bias in data analysis. big data. In many cases, these problems can lead to false results or spurious associations. As an example, studies have falsely correlated chocolate consumption with winning a Nobel Prize or the number of forest fires with shark attacks on California beaches. These two examples are explained by the bias of a third variable that explains the association, the economic level of the country in the first example and high temperatures in the second. On other occasions it is simply chance that generates casual and non-causal associations.

AI systems, although powerful, are not exempt from these challenges. The data quality It is essential to obtain accurate results and prevent errors in analyzes from leading to erroneous interpretations.

Importance of data validation in omics sciences

The omic sciences are a growing field and there is not yet enough verified data to accurately estimate how many studies may be affected by similar errors. However, some audits suggest that between 50% and 89% of published results in preclinical biomedical research are not reproducible. This raises a major concern about the reliability of results in these types of studies.

AI and caution in data analysis

As AI continues to advance in medicine and other disciplines, it is crucial to maintain a critical attitude towards the results generated. Prudence should always be a priority to avoid incorrect interpretations that may arise from the indiscriminate use of Big Data and multiple analyses.

False data can be present in any research, especially in fields such as social sciences and biomedicine. Therefore, it is essential to apply critical thinking and question apparent correlations before accepting them as scientific truths.

For those interested in exploring more about this topic and the implications of biases in statistical analyses, I recommend reading this interesting article from the Cancer Musings blog.

Dr. Ramon Salazar Head of Medical Oncology Service.

ICO L´Hospitalet. Professor of Medicine. UB Campus Bellvitge.