Targeted therapy of tumors and distant metastases of each individual patient is the fundamental task of modern oncology. The choice of treatment regimen, as well as the success rate of therapy, is determined by the individual molecular profile of the tumor. To reduce side effects during therapy, it seems promising to use staged delivery of active agents, or pretargeting: at the first stage, a non-toxic addressable module (antibody, non-immunoglobulin scaffolds) is selectively delivered to a cell of a specific molecular profile, and on a second stage- a cytotoxic agent (toxins, radioactive isotopes) is added. A cytotoxic agent is able to specifically interact with the address module, previously associated with the tumor cell-target,. It is assumed that such a scheme for the introduction of a cytotoxic agent will improve its pharmacokinetics and biodistribution, increase the ratio of tumor / organs, which in turn will reduce the passive absorption of the cytotoxic agent in healthy tissues and thus reduce unwanted side effects. A key problem in the stepwise drug delivery system is the highly accurate in vivo ligation of the addressing module and the cytotoxic agent. Since 1985, to solve this problem 4 main approaches have been developed, based on: 1) the non-covalent interaction between streptavidin and biotin (KD 10^15 M-1); 2) the use of bispecific antibodies capable of interacting with a target on the surface of a tumor cell and with an introduced cytotoxic agent (for example, a “dock-and-lock” system (Rossi E. et al. Proc. Natl. Acad. Sci. USA. 2006, 103, 6841–6846); 3) hybridization of complementary oligonucleotides (Westerlund K. et al. J. Nucl. Med. 2018, 59, 1092-1098); 4) bioorganic chemistry based on tetrazine and trans-cycloethene (Altai M. et al., J. Nucl. Med. 2017, 58, 1553-1559). At the same time, the problem is still far from being solved, since each of these approaches has its limitations (hydrophobicity, high immunogenicity and allergenicity of streptavidin, the presence of endogenous biotin in the body; low binding affinity of components with the nanomolar dissociation constant and biotechnological problems of producing bispecific antibodies (aggregation, problem with solubility), which leads to the high cost of the final product; in the case of synthetic oligonucleotides, the selectivity and speed of hybridization highly dependent on the specific sequence, also has problems with solubility (instability of tetrazine and trans-cycloethene in vivo). Obviously, new solutions and approaches are required. Therefore, the new stepwise delivery technology proposed in this project has a high degree of relevance. In our work, we propose a stepwise drug delivery system in which for the first time the molecular pair barnase-barstar is used for high-precision in vivo binding of the target module and the cytotoxic (or imaging) agent. It is assumed that the address module will be a hybrid protein in wich an anti-HER2-scaffold will be connected with one of the components of the barnase-barstar pair, while the second component of this pair will be associated with a cytotoxic (or imaging) agent. For the pretargeting of HER2-positive tumors, we suggest to use innovative non-immunoglobulin scaffolds - DARPins (Design Ankyrin Repeat Proteins), specifically interacted with different epitopes of the HER2 tumor marker with a dissociation constant in the nanomolar range (Boersma Y. et al. Curr. Opin. Bi. , 849–857) and / or affibody (Löfblom J. et al. Curr. Opin. Biotechnol. 2011, 22, 843–848). Barnase and barstar are two small proteins from Bacillus amyloliquefaciens that can quickly form a strong complex (KD 10-14 M) (Hartley R.W. Methods Enzymol. 2001, 341, 599–611). A significant advantage of the proposed system is a much stronger interaction (by 4-5 orders of magnitude) of the components of the barnase-barstar pair in the second stage of delivery of the cytotoxic agent compared with the interaction of the addressing module with the tumor cell in the first stage of delivery, which ensures fast, strong and highly selective binding of the cytotoxic agent with a target. As a cytotoxic agent, it is supposed to primarily use a fragment of the pseudomonas toxin PE40, which will be used both as part of a hybrid recombinant protein with one of the components of the barnase-barstar pair, and as part of a liposomal construct. Liposomes loaded with cytotoxic protein may be more functionally effective than the “naked” protein, because, firstly, the immunogenicity of the bacterial protein PE40 will be reduced, and secondly, it will be protected from degradation during circulation through the bloodstream. For testing the delivery system, it is desirable to be able to visualize the delivered structure. For this purpose, it is proposed to supplement the cytotoxic agent with a visualizing label, which can be used as infrared fluorescent proteins (for example, iRFP (Filonov G. et al. Nat. Biotechnol. 2011, 29, 757–761), λex / em 690/713 nm) and / or dyes (for example, iRDye800). The scientific novelty of the project is that for the first time in the world in the system of stepwise delivery of a therapeutic agent to a tumor cell, the barnase-barstar molecular pair will be used. This system possesses a number of significant advantages in comparison with the existing approaches. The proposed approach is a new application of the original author's technology, which was previously used to construct multivalent antibodies (Deyev S. et al. Nat. Biotechnol. 2003, 12, 1486-1492).