FGVD: Public Structure Data for Fragment Growing Validation
Bibliographic record
Abstract
Extending ligands and growing fragments is a widely practiced day-to-day modeling technique in structure-based design. A multitude of software packages tackling this task have been developed over the years but many of these are only validated on a hand-picked set of use cases. Common validation techniques include retrospectively finding known actives or prospectively growing new ones and validating these with experiments. There are some attempts to leverage public structural data but these are usually limited to single publications. To facilitate a more systematic and statistically significant approach we propose the Fragment Growing Validation Dataset (FGVD). The FGVD is based on the PDBBind refined set that guarantees a level of structural quality and provides affinity annotation. The FGVD is split into two subsets: the self-growing set and the cross-growing set. In the self-growing set ligands are cut into a core and a fragment inside their own pockets. The cuts are chosen so that the resulting fragments obey the "Rule of Three", the growing directions are roughly towards the protein, and the fragments are not completely solvent exposed. The cross-growing set consists of "growing tasks". A "growing task" is created by aligning two functionally equivalent pockets from the PDBBind refined set and trying to grow the ligand of one structure into the other. In total the self-growing set consists of 4029 test cases across 1096 targets (defined by UniProtKB accessions) using 2035 unique fragments and the cross-growing set consists of 326 test cases accross 93 targets using 155 unique fragments. We characterize the dataset with respect to the phsyico-chemical properties of the fragments, discuss the relevance of the test cases included, as well as provide an example validation using our own fragment growing workflow.
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 | 0.006 | 0.015 |
| Meta-epidemiology (narrow) | 0.004 | 0.002 |
| Meta-epidemiology (broad) | 0.002 | 0.003 |
| Bibliometrics | 0.004 | 0.004 |
| Science and technology studies | 0.002 | 0.001 |
| Scholarly communication | 0.003 | 0.004 |
| Open science | 0.009 | 0.005 |
| Research integrity | 0.006 | 0.005 |
| Insufficient payload (model declined to judge) | 0.025 | 0.033 |
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".