submatrices, the calculated sensitivity and precision value for each submatrix was weighted by the fraction of analyzed elements in a non-reduced matrix (i.e., each calculated metric was then multiplied by the ratio of orange-outlined elements to all elements in the reduced matrix in Fig. 7). Such an approach was only necessary for precision and sensitivity metrics due to the fact that these metrics were calculated on non-overlapping subsets of the initial matrix and their values depended on the size of those subsets, given uneven marker distribution in the initial matrix. Other used metrics (accuracy, specificity) were calculated as mean values between 6 produced metric values (excluding NANs, which originated from calculations involving division by zero).
