OSIELResearch cockpit
OSIEL Research Workspace
DR
OSIEL / DISCOVERYComputational hypothesis

Compound Prioritization Cockpit

Traceable chemical intelligence, uncertainty-aware ranking, and governed experimental learning in one research loop.

Research-use system. Rankings are in-silico prioritization hypotheses—not clinical guidance or measured efficacy.
SCIENTIFIC CONTEXT

Define the governed prediction task

Candidate origin
Task OSIEL-ACTIVITY-NSCLC-v1Split: scaffold-safePolicy: rank-policy@1.0Updated: Reference Analysis
FEDERATED CHEMICAL SPACE10.10B9.43B mapped remotely + 192 local
CANDIDATES IN VIEW85 inside model domain
TOP-5 MEAN SCORE81/100Transparent 7-component policy
EXPERIMENT QUEUE3Selected for computational dry-run
END-TO-END ENGINE

Evidence-to-experiment loop

Ready
01
IdentityStandardize & deduplicate
02
ContextCancer task registry
03
PredictActivity + uncertainty
04
ScreenEndpoint-level ADMET
05
PrioritizePolicy + Pareto fronts
06
ExperimentProtocol dry-run
07
GovernQC + approval gate

Research workspace ready.

CANDIDATE RANKING

Prioritized compounds

#CompoundOriginActivityConfidenceOSIEL scoreDomain
1
QuercetinC15H10O7 · CMP-0001
natural
88%
82% ±18
86.4/100
inside
2
LuteolinC15H10O6 · CMP-0002
natural
84%
79% ±21
83.7/100
inside
3
GefitinibC22H24ClFN4O3 · CMP-0005
synthetic
91%
88% ±12
81.2/100
inside
4
CurcuminC21H20O6 · CMP-0003
natural
80%
73% ±27
78.9/100
borderline
5
CatechinC15H14O6 · CMP-0004
natural
71%
77% ±23
75.6/100
inside
6
5-FluorouracilC4H3FN2O2 · CMP-0006
synthetic
86%
86% ±14
73.3/100
inside
7
ResveratrolC14H12O3 · CMP-0007
natural
68%
67% ±33
70.8/100
borderline
8
NCI reference 0012C10H10N2O · CMP-0012
reference
61%
54% ±46
63.1/100
outside
Showing 8 ranked local candidates · expand through the live 10.10B remote search Selected: 3
EXPERIMENT ORCHESTRATOR

Computational dose-response dry-run

Simulation only
AssayCellTiter-Glo
Dose range0.01–30 µM
Replicates3 × 8 points
Duration72 hours
2Generate plate map
3Simulate response
4Run QC
5Lock result
3 compounds queuedCreates an auditable software-flow test; it does not perform a physical experiment.
MODEL GOVERNANCE

Champion / challenger gate

CHAMPIONOSIEL Activity Model v0.1Approved reference · 2026-08-19
Online
Evaluation boundary
CHALLENGERAwaiting approved datasetTraining never starts from a UI upload
Locked
  • Scientific result approval required
  • Immutable snapshot and leakage audit
  • Independent evaluation and named reviewer