config.yaml

# path to output directory to store result files
out_dir: 
# column name in the input file indicating the column name containing the protein names/IDs
index_col: "protein_ID" 

## Preprocessing params 
# maximum percentage of missing values per protein
max_allowed_NAs_per_protein: 0.3
# transformation funtion to apply to the data before model fitting. options are log10, log2, and log
log_func_name: "log"

# names of columns in the sample annotation file to be passed as covariates to the autoencoder. 
# leaving this empty results in no covariates passed to the model
cov_used:
    #- "gender"
    #- "PROTEOMICS_BATCH"
    #- "BATCH_RUN"
    #- "INSTRUMENT"

## Grid search outlier injection params
inj_freq: 1e-3
inj_mean: 3
inj_sd: 1.6
seed: 

## Model params
autoencoder_training: True ## when set to false, it returns values after (PCA) initialization without training

n_layers: 1 # number of layers for the autoencoder model
n_epochs: 100 # number of training epochs
lr: 1e-4 # learning rate
batch_size: 

find_q_method: "OHT" # option to find latent space dimension. Options: OHT, gs
init_pca: True # PCA initialization
h_dim: # hidden dimension, only for n_layers>1

## Stat params
pval_dist: 't' # Distribution choice for P-value computation. Options: gaussian, t
pval_adj: 'by' # Multiple testing correction. Options: by, bh
pval_sided: 'two-sided'
pseudocount: 0.01 # for FC computation

## Reporting params
outlier_threshold: 0.1
report_all: True