An implementation of a Poisson-JUMP DDPM inspired by (Chen & Zhou, 2023) to denoise low-coverage RNA-seq data.
For julia version 1.10.3
If you wish to use this code, it is recommended to open a Julia REPL in VScode in order to execute the functions and call them.
This code is still very much in developpement and unoptimized, but suggestions are appreciated!
diffusion_model_binom.jl : The min faile for the DDPM developpement
conditional_mlp.jl: Contains the model architectures
samplings.jl: Contains sampling methods, including binomial thinning
time_embeddings.jl: Contains the time embeddings
utils_DDPM.jl: Utility functions
get_counts.jl: File on how counts were obtained and downsampled from L1000.
Lincs.jl: File containing the data structure for L1000. Written by Sébastien Lemieux and Safia Safa-tahar-henni
ML_utils_LINCS.jl: utility functions
The data files are not available on this repo, but can be potentially transferred if requested