•Investigation of carbon nanofiber effects on a wide array of sublethal endpoints.•Comparison of in vitro and in vivo exposures for M. edulis hemocytes.•Linking subtle product features to differences in gene expression. Manufactured nanomaterials are an ideal test case of the precautionary principle due to their novelty and potential environmental release. In the context of regulation, it is difficult to implement for manufactured nanomaterials as current testing paradigms identify risk late into the production process, slowing down innovation and increasing costs. One proposed concept, namely safe(r)-by-design, is to incorporate risk and hazard assessment into the design process of novel manufactured nanomaterials by identifying risks early. When investigating the manufacturing process for nanomaterials, differences between products will be very similar along key physicochemical properties and biological endpoints at the individual level may not be sensitive enough to detect differences whereas lower levels of biological organization may be able to detect these variations. In this sense, the present study used a transcriptomic approach on Mytilus edulis hemocytes following an in vitro and in vivo exposure to three carbon nanofibers created using different production methods. Integrative modeling was used to identify if gene expression could be in linked to physicochemical features. The results suggested that gene expression was more strongly associated with the carbon structure of the nanofibers than chemical purity. With respect to the in vitro/in vivo relationship, results suggested an inverse relationship in how the physicochemical impact gene expression.