Fig. 5
From: Benchmarking principal component analysis for large-scale single-cell RNA-sequencing
(a) Comparison of all combinations of eigenvectors. Absolute values of the cross products of all combinations between the eigenvectors of the gold standard methods and those of the other principal component analysis (PCA) implementations were calculated. The closer the value is to 1 (yellow), the closer the two corresponding eigenvectors are to each other. If two PCA results are equal without considering differences in sign, the matrix in this figure becomes an identity matrix. (b) Comparison of all combinations of loading vectors. Number of intersection elements between the top 500 largest absolute value elements of the gold standard methods and those of the other PCA implementations were calculated. The closer the value is to 500 2 (yellow), the closer the two corresponding loading vectors are to each other. If two PCA results are equal without considering differences in their sign, all the diagonal elements of the matrix in this figure become 500