Thanks very much, @ekoikonomou.bsky.social!

Huge thanks to everyone involved, especially co-first author Ashley Pournamdari and mentors suprêmes
@mchonig.bsky.social and @pnatarajanmd.bsky.social, as well as @broadinstitute.org, @jimjanuzzi.bsky.social, @zhiyu.bsky.social, and everyone else who isn't on Bluesky (yet)!

These results were externally verified in the Women's Health Initiative, corroborating the notion that circulating proteins can significantly improve the prediction of different heart diseases (9/x)

Lastly, we constructed and tested protein-based prediction models for all four cardiac conditions under study.

For CAD, HF, and AF, proteomic models were associated with increments in AUC-ROC ranging from 0.02 to 0.06. There was no significant improvement for AS (8/x)

Moving back to the traditional epidemiological analyses, we also identified some proteins with evidence for between-sex differences (7/x)

Proteogenomic analyses - including MR and colocalization analyses - prioritized a subset of identified proteins (~4%) as therapeutic targets.

Examples with strong evidence include ADM and SPON1 for AF and SPINT1 for CAD (6/x)

Approximately 60% of identified proteins were associated with more than one cardiac disease; however, we also identified biomarkers specific to one condition.

Similarly, enrichment analyses identified several shared and distinct protein pathways for the four diseases (5/x)

Primary analyses - using multivariable-adjusted Cox regression models and time-varying covariates - identified 820 Bonferroni-significant protein-disease associations.

NT-proBNP, GDF15, and WFDC2 were among the top hits (4/x)

Using data from 44,313 participants from the
UK Biobank, we tested the associations of 1,459 circulating proteins with these four cardiac diseases (3/x)

Cardiac diseases such as coronary artery disease (CAD), heart failure (HF), atrial fibrillation (AF), and aortic stenosis (AS) account for the majority of cardiac deaths.

Yet, underlying mechanisms not captured by traditional risk factors remain incompletely understood (2/x)