Fig. 4: Performing a pooled SARS-CoV-2 RT-LAMP workflow using a ferrobot swarm.

a, Schematic of the square matrix pooling scheme. The flow chart at the centre provides an overview of the infected sample identification process based on the assay pooled (A) or row–column (R_i/C_j) responses. b, Sequential optical images of an automated 4² pooling workflow performed by a team of nine ferrobots. To combine aliquots in each pooling step, they were ferrobotically collected, merged, mixed and then dispensed as a 1-μl droplet. The inset images show the critical intermediary ferrobotic operations. c,d, Optical images and readouts obtained from ferrobotic pooled testing of two groups of nine clinical samples using the 3² pooling chip. The negative assay A response indicated that no infected sample was present among the first group of samples in c. The positive assay A response along with the positive assay R₂ and C₂ responses led to the identification of the infected sample (located at the second row–second column) among the second group of samples (d). e,f, Optical images and readouts obtained from ferrobotic pooled testing of two groups of 16 clinical samples using the 4² pooling chip. The negative assay A response indicated that no infected sample was present among the first group of samples in e. The positive assay A response along with the positive assay R₃ and C₃ responses led to the identification of the infected sample (located at the third row–third column) among the second group of samples (f). In c–f, error bars indicate different trials of optical reading, mean ± s.e. (n = 5). Horizontal dashed line indicates on-chip detection threshold (710 a.u.).