Nilus Lab is a computational biotechnology company combining 12 public cardiac datasets (11 human, 1 mouse) with human genetics and mouse knockout data to identify candidate targets for heart failure and myocardial ageing. All candidates are at the computational identification stage.
Single-cohort transcriptomic screens in heart failure frequently fail because they ignore non-myocyte vascular and stromal niches, miss post-transcriptional protein accumulation in long-lived cardiomyocytes, and mistake compensatory protective stress responses for pathogenic drivers.
To overcome these failure modes, our discovery engine integrates 6 human single-cell/single-nucleus atlases (Reichart 2022, PERIHEART, CAREBANK, Kanemaru, Koenig, Litviňuková), 3 bulk RNA-seq cohorts (MAGNet, GTEx LV/AA), 3 mass-spectrometry cardiac proteomes (Malecki 2025 human LV, 12 young vs 13 older donors; GTEx human heart, 7 LV and 11 atrial donors; Takasugi 2024 mouse heart), and Open Targets v4 human genetics.
Without prior gene weighting, unblinded decomposition recovered the clinical gene-therapy benchmark ATP2A2 (SERCA2a) blind, ranked 76th of 5,677 genes and 3rd of the 11 genes left after all filters ($Z_{\text{DCM}} = -7.994$, Malecki human LV protein $p = 0.00039$, $q = 0.066$). It also produced three computational candidates: PPP1R12C (MBS85) (higher in older human LV at protein level, $p = 1.1\times10^{-4}$; passes FDR in the main test only; its first intron holds the AAVS1 locus), TMEM65 (strongest heart-failure signal of the three, $Z = -8.008$) and SYPL2 (MG29) (lower in failing cardiomyocytes; its genetic link is a liver eQTL at a shared locus, so it is not counted as heart evidence). None changes with age in GTEx left ventricle, and none has been tested in the lab. The audit also filtered out universal systemic non-cardiac ageing markers (ZMAT3, LTBP2) and compensatory protective brakes (ARHGAP1, NOL3, CIRBP).
Combines non-myocyte single-cell pseudobulk regressions (with raw-count ambient soup filtering), dual-species mass-spectrometry proteomics, DCM vs. ageing decomposition across 451 FDR-concordant genes, and GWAS colocalisation ($H_4 \ge 0.80$).
Audits every candidate against HAGR GenAge/CellAge, EuropePMC literature counts, and IMPC mouse knockout phenotypes — catching and rejecting compensatory protective brakes (ARHGAP1, NOL3/ARC, CIRBP).