ClarusC64/ABX-PD-004_mutant_prevention_window_collapse-v0.1
ABX-PD-004: Mutant Prevention Window Collapse This dataset tests early resistance risk before MIC moves. The signal is a rising MPC that makes the mutant prevention window widen fast. What you track MIC in mg/L MPC in mg/L exposure_max_mg_L as a proxy for achievable concentration Derived values mpw_ratio = MPC / MIC mpc_cover_margin = exposure_max_mg_L / MPC Labels mpw_collapse_signal 1 for rows at or after collapse earliest_collapse 1 only… See the full description on the dataset page: https://huggingface.co/datasets/ClarusC64/ABX-PD-004_mutant_prevention_window_collapse-v0.1.
ABX-PD-004: Mutant Prevention Window Collapse
This dataset tests early resistance risk before MIC moves.
The signal is a rising MPC that makes the mutant prevention window widen fast.
What you track
- MIC in mg/L
- MPC in mg/L
- exposuremaxmg_L as a proxy for achievable concentration
Derived values
- mpw_ratio = MPC / MIC
- mpccovermargin = exposuremaxmg_L / MPC
Labels
- mpwcollapsesignal
- 1 for rows at or after collapse
- earliest_collapse
- 1 only for the first detected row in that series
v1 detection idea
You call collapse when
- MIC stays similar
- MPC rises sharply
- MPC becomes unachievable
- mpccovermargin < 1
- or the MPW becomes very wide
- mpw_ratio >= 16
The scorer also flags spike and snap MPC values as likely measurement artifacts.
Files
- data/train.csv
- data/test.csv
- scorer.py
Evaluation
Run
- python scorer.py --path data/test.csv
The scorer reports
- series level F1 for mpwcollapsesignal
- earliest collapse hits and misses
- exposure drop flags
- measurement spike flags