Transfer learning efficiently maps bone marrow cell types from mouse to human using single-cell RNA sequencing.

Biomedical research often involves conducting experiments on model organisms in the anticipation that the biology learnt will transfer to humans. Previous comparative studies of mouse and human tissues were limited by the use of bulk-cell material. Here we show that transfer learning-the branch of machine learning that concerns passing information from one domain to another-can be used to efficiently map bone marrow biology between species, using data obtained from single-cell RNA sequencing. We first trained a multiclass logistic regression model to recognize different cell types in mouse bone marrow achieving equivalent performance to more complex artificial neural networks. Furthermore, it was able to identify individual human bone marrow cells with 83% overall accuracy. However, some human cell types were not easily identified, indicating important differences in biology. When re-training the mouse classifier using data from human, less than 10 human cells of a given type were needed to accurately learn its representation. In some cases, human cell identities could be inferred directly from the mouse classifier via zero-shot learning. These results show how simple machine learning models can be used to reconstruct complex biology from limited data, with broad implications for biomedical research.

Transcutaneous immunization with a highly active form of XCL1 as a vaccine adjuvant using a hydrophilic gel patch elicits long-term CD8+ T cell responses.

Memory CD8+ cytotoxic T-lymphocytes (CTLs) play a key role in protective immunity against infection and cancer. However, the induction of memory CTLs with currently available vaccines remains difficult. The chemokine receptor XCR1 is predominantly expressed on CD103+ cross-presenting dendritic cells (DCs). Recently, we have demonstrated that a high activity form of murine lymphotactin/XCL1 (mXCL1-V21C/A59C), a ligand of XCR1, can induce antigen-specific memory CTLs by increasing the accumulation of CD103+ DCs in the vaccination site and the regional lymph nodes. Here, we combined a hydrophilic gel patch as a transcutaneous delivery device and mXCL1-V21C/A59C as an adjuvant to further enhance memory CTL responses. The transcutaneous delivery of ovalbumin (OVA) and mXCL1-V21C/A59C by the hydrophilic gel patch increased CD103+ DCs in the vaccination site and the regional lymph nodes for a prolonged period of time compared with the intradermal injection of OVA and mXCL1-V21C/A59C. Furthermore, the hydrophilic gel patch containing OVA and mXCL1-V21C/A59C strongly induced OVA-specific memory CTLs and efficiently inhibited the growth of OVA-expressing tumors more than the intradermal injection of OVA and mXCL1-V21C/A59C. Collectively, this type of hydrophilic gel patch and a high activity form of XCL1 may provide a useful tool for the induction of memory CTL responses.

Marked load-bearing ability of Mytilus smooth muscle in both active and catch states as revealed by quick increases in load.

The anterior byssal retractor muscle (ABRM) of the bivalve Mytilus edulis shows a prolonged tonic contraction, called the catch state. To investigate the catch mechanism, details of which still remain obscure, we studied the mechanical responses of ABRM fibres to quick increases in load applied during maximum active isometric force (P(0)) generation and during the catch state. The mechanical response consisted of three components: (1) initial extension of the series elastic component (SEC), (2) early isotonic fibre lengthening with decreasing velocity, and (3) late steady isotonic fibre lengthening. The ABRM fibres could bear extremely large loads up to 10-15P(0) for more than 30-60 s, while being lengthened extremely slowly. If, on the other hand, quick increases in load were applied during the early isometric force development, the ABRM fibres were lengthened rapidly ('give') under loads of 1.5-2P(0). These findings might possibly be explained by two independent systems acting in parallel with each other; one is the actomyosin system producing active shortening and active force generation, while the other is the load-bearing system responsible for the extremely marked load-bearing ability as well as the maintenance of the catch state.