Abstract B036: Machine learning-based and marker-based combined strategy for circulating tumor cells candidates' detection
Bibliographic record
Abstract
Abstract Introduction: The identification of circulating tumor cells (CTCs) in single-cell transcriptomic datasets is of considerable clinical and biological importance, as CTCs are critical in the metastatic cascade and serve as potential biomarkers for cancer prognosis and therapeutic monitoring. However, their detection remains challenging due to their extreme rarity and phenotypic fluidity, partially overlapping with the phenotype of Peripheral Blood Mononuclear Cells (PBMCs) and other cells found in the bloodstream. Materials and Methods: To address the classification uncertainty of individual CTCs, we combined our previously developed machine learning (ML) models with a marker-based strategy. We utilized a publicly available dataset (GEO GSE109761), which included 262 annotated CTCs, 14 CTC–PBMC clusters (treated as CTCs), and 82 PBMCs. Our most effective ML model used only 67 features and achieved a balanced accuracy of 0.95 (in differentiating between CTCs and other cells) on a test set comprising 132 CTCs, 43 WBCs, and 6 CTC–PBMC clusters. Subsequently, we applied this model to a single-cell dataset derived from breast cancer patients, which included 6 patients with triple-negative breast cancer (TNBC), 2 with Luminal B HER2-negative, 2 with Luminal B HER2-positive, and 3 control patients with non-malignant conditions. Given the unexpectedly high number of predicted CTCs (up to 600 cells per sample), we refined our selection using a marker-based filtering approach. From the GEO dataset, we selected CTCs exhibiting epithelial or mesenchymal phenotypes and compared them with PBMCs to identify the 100 most discriminative genes. These markers were then used to score cells in the breast cancer dataset using the UCell R package. Cells lacking leukocyte-associated scores but showing epithelial or mesenchymal scores exceeding those found in control samples were retained as candidate CTCs. Results: We identified 5 CTC candidates with epithelial phenotype, 2 with mesenchymal phenotype, and 5 with an intermediate phenotype. Discussion: While definitive classification of cells as CTCs cannot be confirmed, our integrative strategy, combining robust ML predictions with a stringent marker-based refinement, enhances the confidence in their identification. Citation Format: Michał Sieczczyński, Krzysztof Pastuszak, Anna Supernat, Anna J. Żaczek. Machine learning-based and marker-based combined strategy for circulating tumor cells candidates' detection [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Artificial Intelligence and Machine Learning; 2025 Jul 10-12; Montreal, QC, Canada. Philadelphia (PA): AACR; Clin Cancer Res 2025;31(13_Suppl):Abstract nr B036.
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 | 0.003 | 0.004 |
| Meta-epidemiology (narrow) | 0.002 | 0.000 |
| Meta-epidemiology (broad) | 0.002 | 0.002 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.002 | 0.001 |
| Research integrity | 0.002 | 0.002 |
| Insufficient payload (model declined to judge) | 0.004 | 0.002 |
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".