Danil D. Kotelnikov

• 

  NanoDeNovo: nanobodies designed from the antigen alone

  I built a five-stage pipeline that designs nanobodies against the poliovirus type 1 VP3 capsid protein. No immunisation, no display library. It samples scaffolds, predicts their structures, docks them to the antigen, redesigns the CDR loops on whatever survives, and humanises the final candidates. Rosetta relaxation runs last because it costs about a thousand times more per candidate than folding does. The five stages use different models and different failure criteria, so their errors have to be checked separately.

  Paper Method note

  Fig. 8. Kotelnikov, Tatarinova, Zhdanov, 2025. doi:10.3390/ijms26199262. CC BY 4.0.
• 

  Non-coding RNA and the eusocial transition

  I reviewed the evidence linking regulatory RNAs to caste development and social behaviour in bees, ants and wasps. The paper maps each RNA class to its mechanism and to the genes it regulates. It then asks how far expression profiles follow the shift from facultative to obligate sociality, and whether they help explain why that shift is hard to reverse.

  Paper

  Fig. 1. Lebedev, Smutin, Timkin, Kotelnikov, Taldaev, Panushev, Adonin, 2025. doi:10.1016/j.ncrna.2024.10.007. CC BY 4.0.

• Primery: primers placed on mismatch clusters

  Primery finds species-specific regions before any primer is designed. It BLASTs the target against its relatives, groups the discriminating positions into primer-length windows, and scores each window by how many species it separates and how consistently. Primer3 then designs inside the best windows under the constraints of the chosen chemistry. Designed pairs go back to BLAST for amplicon classification. When a locus cannot separate species, the report says so and gives the genus-level result.

  Method note The clustering argument

  Top cluster discrimination score per benchmark locus. Manuscript in preparation.

• Metal complexes as biological agents: structure and binding

  Gold(III) complexes with doubly protonated phenanthrolines, and binuclear copper(II) furancarboxylates with 5-nitro-1,10-phenanthroline. X-ray diffraction gave the copper coordination environment as a square pyramid, coordination number five, held together by hydrogen bonds and stacking between the aromatic rings. I worked on the computational side, modelling how Cu²⁺ binds mycobacterial proteins. Histidine and glutamate sites dominated. Both copper complexes suppressed viability in an ovarian adenocarcinoma line.

  Gold(III) paper Copper(II) paper

  Neither article is openly licensed, so no figure is reproduced here.
• 

  Cross-checked ligand docking on an ion channel

  We screened a panel of synthetic glucocorticoids against a predicted TRPM8 structure, taken from a structure database because no crystal structure was available. Each pose was computed twice, once in AutoDock and once in MOE. We then compared the two interaction maps residue by residue and accepted a result only where they agreed. One compound reached the residue set we cared about. The others formed stable complexes elsewhere on the channel.

  Paper

  Ligand interaction diagrams. Timkin, Kotelnikov, Timofeev, Naumov, Borodin, 2024. doi:10.20538/1682-0363-2024-4-136-144. CC BY 4.0.

• Aug 2026 Primery: turning BLAST mismatches into species-specific primers Stage parameters, the scoring function, a benchmark across 21 loci, and how to take a designed pair to the bench.
• Jul 2026 Placing a primer on a mismatch cluster Computing cluster density from an alignment and converting the melting-temperature gap into a gradient window.
• May 2026 Assembling a plant genome from short reads Profiling order, three assembler architectures compared, and how to test a hybrid assembly for chimerism.

All notes