Plasma Proteomes Can Be Reidentifiable and Potentially Contain Personally Sensitive and Incidental Findings

Abstract

The goal of clinical proteomics is to identify, quantify, and characterize proteins in body fluids or tissue to assist diagnosis, prognosis, and treatment of patients. In this way, it is similar to more mature omics technologies, such as genomics, that are increasingly applied in biomedicine. We argue that, similar to those fields, proteomics also faces ethical issues related to the kinds of information that is inherently obtained through sample measurement, although their acquisition was not the primary purpose. Specifically, we demonstrate the potential to identify individuals both by their characteristic, individual-specific protein levels and by variant peptides reporting on coding single nucleotide polymorphisms. Furthermore, it is in the nature of blood plasma proteomics profiling that it broadly reports on the health status of an individual—beyond the disease under investigation. Finally, we show that private and potentially sensitive information, such as ethnicity and pregnancy status, can increasingly be derived from proteomics data. Although this is potentially valuable not only to the individual, but also for biomedical research, it raises ethical questions similar to the incidental findings obtained through other omics technologies. We here introduce the necessity of—and argue for the desirability for—ethical and human-rights-related issues to be discussed within the proteomics community. Those thoughts are more fully developed in our accompanying manuscript. Appreciation and discussion of ethical aspects of proteomic research will allow for deeper, better-informed, more diverse, and, most importantly, wiser guidelines for clinical proteomics.

Artificial intelligence for proteomics and biomarker discovery

“…That said, we have noted an increasing tension between the imperatives for data accessibility, as formalized in the FAIR principles (Wilkinson et al, 2016), and the desire for data privacy, such as embodied in the GDPR regulation of the European Union (Ducato, 2020) (Figure 5). This has only recently become an issue in proteomics as data quality has increased to an extent that it potentially allows (re)identification of individuals, which now raises ethical and philosophical questions (Geyer et al, 2021;Mann et al, 2021). It is important that the different stakeholders and authorities recognize these challenges and deal with them in a way that does not prevent scientific and medical progress.…”

Forensic proteomics

“…Protein contains genetic information in the form of single amino acid polymorphisms (SAPs), the result of non-synonymous SNPs [43,[241][242][243]. Proteomic genotyping is the detection of genetically variant peptides (GVPs) that contain SAPs and then inferring the presence of the corresponding SNP alleles in the genome of the individual who produced the protein sample (Figure 3).…”

“…This includes telogenic hair shafts, degraded and compromised bones and teeth, fingermarks, and sexual assault evidence. The method however can be used for any protein source [9,242].…”

The growing need for controlled data access models in clinical proteomics and metabolomics

“…High-throughput proteomics platforms have rapidly evolved in recent years and now have the potential to collect sufficient data to make single individual personally identifiable 5 . Currently, potentially sensitive proteomics datasets of clinical interest are routinely made openly available via the PX resources, but ethical discussions over clinical proteomics data have begun recently 5,6 . At PRIDE, we are receiving an increasing number of queries about data management of sensitive human data.…”

“…The highest risk to render individuals identifiable by proteomics lies in the MS raw data and the amino acid sequence database. In addition, re-identification of individuals was recently reported based on protein expression information from plasma samples 5 , even without variant sequence information at the protein level. In a recent community-driven paper 6 , the potential of different proteomics data types to provide patient identifiable information was discussed.…”