Transcriptome similarity explained a small part of residual coupling in 40 off-food and 8 re-fed worms

We tested whether molecular neuron properties explain reliable coupling left after measured wiring and broad cell controls. We used a fixed similarity of 13,669 log-TPM gene profiles from CeNGEN and activity from 40 off-food and 8 re-fed C. elegans. We excluded pairs assigned the same expression profile. Pair models removed additive neuron effects, distance and axis-wise soma geometry, chemical and electrical wiring, shared chemical inputs, and broad type-by-transmitter blocks. Transcriptome similarity had a positive coefficient before and after out-of-fold behaviour and drift removal in both cohorts. All four planned within-identity expression-class permutation tests had Holm-adjusted p = 0.012. After behaviour and drift removal, estimated molecular shares of reliable remaining coupling were 3.46 percent (95 percent animal jackknife t interval 2.63 to 4.29) and 2.43 percent (-0.32 to 5.18). No single expression class drove the link. It remained in pairs with no direct link and in a first-800-frame window before heat delivery. On 3,174 common pairs, fixed coefficients transferred with positive point gains, but one residual-transfer interval included zero. Wiring and molecular terms together explained only 6.6 to 7.5 percent of reliable residual coupling after a common identity and geometry baseline. The result identifies a small molecular correlate. It does not identify a causal gene mechanism. Different source animals, class-average larval expression, mixed recording protocols, one laboratory, an eight-animal replication, unmeasured drive, and a limited wiring model constrain the result. Code, checked source hashes, aggregate evidence, figures, and an external-data retrieval recipe are included; external records are excluded.