GABI

Genome Annotation using Bayesian Inference (GABI)

How to use:

Simple Core Version

Inputs:

• matrix: MxN binary matrix
• labels: M dim vector of profiles clusters (or cell type) membership

Output:

• matrixCT: (CxN binary matrix with C clusters (or cell types) ) Cleaned matrix
• matrixCTProba: (CxN binary matrix with C clusters (or cell types) ) Cleaned matrix probability to have a peak
• combmat: (CxK binary matrix with K combinations) Combinations between the cell types
• statemat: (MxK binary matrix with K combinations) Combinations between the cell types with all samples
• labels_states: (N integer vector) combinations position along the genome

import GABI as gbi

gb = gbi.GABI(labels)
gb.fit(matrix)
matrixCT = gb.predict(matrix)

#Additional
labels_states = gb.labels_state
statemat = gb.statemat
matrixCTProba = gb.predict(matrixC,GetProba=True)

Multiprocessing Version

The multiprocessing version slice the samples in groups of similar cell types and apply GABI on each part before merging them.

Additional inputs:

• distmat: (MxM distance matrix) Distance between the profiles. If empty, yule distance is used.
• Nclust: Number of cell types per slices

import GABI_MP as gbimp

gb = gbimp.GABI(matrix,labels,distmat=[],verbose=True,NClust=8)
gb.fit()
matrixCT,statemat,labels_states = gb.predict()

#Additional:
combmat = gb.combmat

Note

KeyboardInterrupt is handled in both version if we want to end the fit before the end.