Fig. 2

Overview of IMLARF and an example of LARF. a Overview of IMLARF. The algorithm consists of three steps, the construction of matrix (i) M and (ii) F and (iii) pairwise product of M and F. In ISLARF, the matrix M in step 1 is simply replaced with the matrix S (Section “Statistical approach”). b An example of procedures of LARF. It shows how the row vector F ¹ of frequency matrix F given target gene G1 and 8 other candidate regulators (G2 ∼G9). By a predefined α, four random features are selected among eight genes in each iteration. In the beginning, F ¹ is not increased and four random features are selected without sparsity since λ is not increased enough yet. The more λ is increased, the more the number of selected features (blue-colored cells) is decreased. If no feature is selected due to a highly increased λ, the iteration and frequency measure is finished