The application of wearable magnetoencephalography using optically-pumped magnetometers has drawn extensive attention in the field of neuroscience. Electroencephalogram system can cover the whole head and reflect the overall activity of a large number of neurons. The efficacy of optically-pumped magnetometer in detecting event-related components can be validated through electroencephalogram results. Multivariate pattern analysis is capable of tracking the evolution of neurocognitive processes over time. In this paper, we adopted a classical Chinese semantic congruity paradigm and separately collected electroencephalogram and optically-pumped magnetometer signals. Then, we verified the consistency of optically-pumped magnetometer and electroencephalogram in detecting N400 using mutual information index. Multivariate pattern analysis revealed the difference in decoding performance of these two modalities, which can be further validated by dynamic/stable coding analysis on the temporal generalization matrix. The results from searchlight analysis provided a neural basis for this dissimilarity at the magnetoencephalography source level and the electroencephalogram sensor level. This study opens a new avenue for investigating the brain\'s coding patterns using wearable magnetoencephalography and reveals the differences in sensitivity between the two modalities in reflecting neuron representation patterns.