Capturing Genetic Diversity in Proteomics

When genome sequencing first became available, the field established the “reference” human genome – an arbitrary reference sequence used to map the differences that make us who we are. The field of precision medicine aims to take these differences into account to better personalize treatments – but not all populations are equally represented in medical studies which can create biases. In genomics, for example, clinical usage of established associations between genetic variants and medical conditions (polygenic risk scores) may increase health disparities for underserved communities.

When studying large human cohorts in proteomics, the data are aligned onto the reference proteome, the product of the reference genome. All natural and common differences between humans are lost. Furthermore, it is easy to mistake two sequences that are similar. If we only search for reference sequences, some parts of proteins that are varying would be falsely matched, while others will remain invisible.

Accounting for the differences between humans opens the possibility of avoiding such errors and accounting for genetic diversity in medical research better.

This work is part of a broader project funded by the Norwegian Research Council to enable the interpretation of proteomic data in the context of genetic variation. Proteomic data contains the products of genetic variation – we were just blind to it so far. To be able to chart variation in human samples, we adopted the methods used in genetic epidemiology and wrote ProHap to produce a map of so-called protein haplotypes, that would fairly represent the participants of reference genetic panels.

First, we have used ProHap with the genotypes of the 1000 Genomes Project to generate 6 different databases of protein sequences. In the 1,000 Genomes Project, participants were pooled into 5 groups:  African, American, European, East Asian and South Asian. The first 5 databases therefore represent these groups, while the 6^th contains protein sequences commonly expected among all the individuals from the 1000 Genomes panel.

The 1000 Genomes are not a perfect sample of human diversity, but they do offer a glimpse into the differences between the reference proteome and individuals in the different populations. We have seen that the highest share of the proteome can be affected by genetic variation in the African “superpopulation”, while all individuals would benefit from having their genotypes accounted for in proteomic studies.

We have also used ProHap to create a personalized proteome for a donor whose genome sequence is publicly available online. Stem cells derived from this donor are commercially available and, when analyzing them, we found that we can detect many changes in the protein sequences. In some cases where the donor carries two different versions of the same gene, we can detect the two different proteins that can be encoded.

Firstly, the proteomics community can readily use the sequence databases that we have published alongside the tool. These databases will be useful for the development of more refined proteomic workflows.

Moreover, as the field of genomics moves to produce various new panels of genotypes, with this data rightfully staying in the ownership of the respective communities, ProHap can be used on secure servers to produce new protein databases, maintaining the confidentiality of the data.

Finally, we are investigating the usage of databases created by ProHap in studies of the immune system, where refining the protein analyses to the personal level is vital (e.g., when studying graft–host interaction).