Identification of Potential PBT/POP-Like Chemicals by a Deep Learning Approach Based on 2D Structural Features
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
Identifying potential persistent organic pollutants (POPs) and persistent, bioaccumulative, and toxic (PBT) substances from industrial chemical inventories are essential for chemical risk assessment, management, and pollution control. Inspired by the connections between chemical structures and their properties, a deep convolutional neural network (DCNN) model was developed to screen potential PBT/POP-like chemicals. For each chemical, a two-dimensional molecular descriptor representation matrix based on 2424 molecular descriptors was used as the model input. The DCNN model was trained via a supervised learning algorithm with 1306 PBT/POP-like chemicals and 9990 chemicals currently known as non-POPs/PBTs. The model can achieve an average prediction accuracy of 95.3 ± 0.6% and an F-measurement of 79.3 ± 2.5% for PBT/POP-like chemicals (positive samples only) on external data sets. The DCNN model was further evaluated with 52 experimentally determined PBT chemicals in the REACH PBT assessment list and correctly recognized 47 chemicals as PBT/non-PBT chemicals. The DCNN model yielded a total of 4011 suspected PBT/POP like chemicals from 58 079 chemicals merged from five published industrial chemical lists. The proportions of PBT/POP-like substances in the chemical inventories were 6.9-7.8%, higher than a previous estimate of 3-5%. Although additional PBT/POP chemicals were identified, no new family of PBT/POP-like chemicals was observed.
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.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
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 it