Toxic Colors: The Use of Deep Learning for Predicting Toxicity of Compounds Merely from Their Graphic Images
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
The majority of computational methods for predicting toxicity of chemicals are typically based on "nonmechanistic" cheminformatics solutions, relying on an arsenal of QSAR descriptors, often vaguely associated with chemical structures, and typically employing "black-box" mathematical algorithms. Nonetheless, such machine learning models, while having lower generalization capacity and interpretability, typically achieve a very high accuracy in predicting various toxicity endpoints, as unambiguously reflected by the results of the recent Tox21 competition. In the current study, we capitalize on the power of modern AI to predict Tox21 benchmark data using merely simple 2D drawings of chemicals, without employing any chemical descriptors. In particular, we have processed rather trivial 2D sketches of molecules with a supervised 2D convolutional neural network (2DConvNet) and demonstrated that the modern image recognition technology results in prediction accuracies comparable to the state-of-the-art cheminformatics tools. Furthermore, the performance of the image-based 2DConvNet model was comparatively evaluated on an external set of compounds from the Prestwick chemical library and resulted in experimental identification of significant and previously unreported antiandrogen potentials for several well-established generic drugs.
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 | 0.001 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.002 |
| Open science | 0.000 | 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