Evaluating Virtual Screening Methods: Good and Bad Metrics for the “Early Recognition” Problem
Bibliographic record
Abstract
Many metrics are currently used to evaluate the performance of ranking methods in virtual screening (VS), for instance, the area under the receiver operating characteristic curve (ROC), the area under the accumulation curve (AUAC), the average rank of actives, the enrichment factor (EF), and the robust initial enhancement (RIE) proposed by Sheridan et al. In this work, we show that the ROC, the AUAC, and the average rank metrics have the same inappropriate behaviors that make them poor metrics for comparing VS methods whose purpose is to rank actives early in an ordered list (the "early recognition problem"). In doing so, we derive mathematical formulas that relate those metrics together. Moreover, we show that the EF metric is not sensitive to ranking performance before and after the cutoff. Instead, we formally generalize the ROC metric to the early recognition problem which leads us to propose a novel metric called the Boltzmann-enhanced discrimination of receiver operating characteristic that turns out to contain the discrimination power of the RIE metric but incorporates the statistical significance from ROC and its well-behaved boundaries. Finally, two major sources of errors, namely, the statistical error and the "saturation effects", are examined. This leads to practical recommendations for the number of actives, the number of inactives, and the "early recognition" importance parameter that one should use when comparing ranking methods. Although this work is applied specifically to VS, it is general and can be used to analyze any method that needs to segregate actives toward the front of a rank-ordered list.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.046 | 0.157 |
| Meta-epidemiology (narrow) | 0.003 | 0.001 |
| Meta-epidemiology (broad) | 0.004 | 0.002 |
| Bibliometrics | 0.007 | 0.003 |
| Science and technology studies | 0.001 | 0.005 |
| Scholarly communication | 0.005 | 0.007 |
| Open science | 0.003 | 0.004 |
| Research integrity | 0.004 | 0.004 |
| Insufficient payload (model declined to judge) | 0.001 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".