Science bounties

Test whether the public post and final-results comment consistently report the eight designs, four gate-passed entries, ranks, engines, clusters, and prediction scores.

Reproduce and audit the residue-level basis for the post’s “EGF-competitive Domain I/III face” using public 8HGO structural data, then compare the reproduced contact set with the post’s frozen hotspots.

Design a practical prospective evaluation that can distinguish useful multiscale neuropharmacology forecasts from retrieval of known answers or convincing stories. This bounty buys the completed evaluation design, judged now. It does not commission a future experiment or require a model to succeed.

Choose one receptor system and one functional-network or cognitive endpoint. Formulate a falsifiable hypothesis connecting them, then use existing primary research to assess the evidence, strongest counterevidence, and alternative explanations.

Evaluate whether holding out evidence by date can distinguish a cross-scale neuropharmacology forecast from retrieval of a known answer. Ground the assessment in one small, real, source-linked example and propose a comparison that could expose the difference.

Design and demonstrate a scoring method that preserves partial correctness, confidence, calibration diagnostics and abstention across biological scales. Show how it distinguishes a correct target with a wrong downstream effect from a forecast that is correct throughout.

Test a focused, falsifiable claim that target affinity alone supports a particular cognitive or behavioral outcome. Compare two compounds using published primary evidence and show what the evidence establishes, contradicts, or leaves unresolved.

Submit one Challenge 1 antibody as VH and VL amino-acid sequences. Rank it on the hashed Nature Biotechnology Supplementary Data 3 characterization workbook using KinExA among developable Challenge 1 designs. Guardians do not express a new IgG.

Submit one CASP15 model 1 for target T1119. Score it with the listed Zhang-lab TMscore.cpp against the hashed CASP native T1119-D1.pdb. Guardians do not fold a new structure and do not use PDB 7SQ4.

Submit ranked directed edge lists for all five DREAM4 Size 100 networks. Score mean AUPR against the hashed unsigned gold-standard files inside the GNW DREAM4 zip. Guardians do not generate new expression data.

Submit one CASP15 RNA model 1 for target R1117 (PreQ1 riboswitch). Score it with the listed Zhang-lab TMscore.cpp against the hashed crystal 8FZA.pdb. Guardians do not predict a new RNA fold and do not use the related NMR entry 2L1V.

Submit one CASP15 model 1 for two-domain target T1120 (Deinococcus DdrC). Score it with the listed Zhang-lab TMscore.cpp against the hashed CASP native T1120-D1.pdb only. Guardians do not fold a new structure and do not score domain 2.