Mapping and Deciphering Neural Codes of NMDA Receptor-Dependent Fear Memory Engrams in the Hippocampus
(A) From spike trains of CA1 units, firing rates in two 250 msec windows after stimuli presentation was extracted. (B) After binned and normalized, firing rates were transformed to measure to neural responses. The matrix of normalized response from CA1 units corresponding to the sampling points (all repetitions of CS or US or Rest) was obtained. (C) The covariance matrix can be determined by between-class matrix and within-class matrix, which were obtained from population responses matrix. (D) The discriminant projection vectors are determined by the eigenvalue decomposition of covariance matrix. (E) Transfer matrix was constructed with the corresponding eigenvectors as columns and were sorted in the descend order of the eigenvalues. (F) Neural ensemble responses are then projected to form event- and resting state- clusters in MDA pattern encoding subspaces by transfer matrix. The top three most discriminant subspaces (MDA1-3) are plotted for intuitive visualization. A sliding-window technique can be further applied to calculate transient ensemble states of neural activity (using 20 millisecond steps), thereby tracking dynamic evolution of ensemble trajectories in time throughout the entire recording experiments.