DeCovarT - Covariance-Aware Deconvolution of Bulk Transcriptomic Samples

Estimates cell-type proportions in bulk transcriptomic samples with a probabilistic convolution model that integrates the gene-gene covariance structure of purified reference profiles. Cellular ratios are recovered by maximum likelihood under a multivariate Gaussian convolution, using analytic gradients and Hessians, an isometric log-ratio reparametrisation (Helmert basis) that enforces the simplex constraint, and Marquardt-Levenberg or Newton-type optimisers. The methodology is described in Chassagnol, Nuel and Becht (2023) <doi:10.48550/arXiv.2309.09557>.