MétaCan
Menu
← Back to cohort

Abstract A032: Using machine learning to tackle tumor heterogeneity

2025· article· en· W4412163844 on OpenAlexaboutno aff
Jennifer Neumaier, Luisa Bresadola, Jonas Ibn-Salem, Ranganath Gudimella, Pablo Riesgo Ferreiro, Barbara Schrörs, Uğur Şahin

Bibliographic record

VenueClinical Cancer Research · 2025
Typearticle
Languageen
FieldMedicine
TopicRadiomics and Machine Learning in Medical Imaging
Canadian institutionsnot available
Fundersnot available
KeywordsTumor heterogeneityCancerMedicineComputer scienceInternal medicine

Abstract

fetched live from OpenAlex

Abstract Cancer develops through an evolutionary process, which creates highly diverse primary tumors and metastases. This intra- and intertumoral heterogeneity is a challenging aspect for clinical practice. Selecting therapeutic targets that exist only in a subpopulation of tumor cells can lead to ineffective treatments and in worst cases even to more resistant tumors and relapses. Therefore, the characterization of tumor heterogeneity and the ability to identify targets that are present in all cells of a primary tumor and related metastases are essential for an effective therapy. Truncal mutations are somatic mutations that appear very early in tumor development and are therefore carried by all cells of a tumor throughout subsequent generations. Currently, multiple tumor samples from a patient are necessary to identify truncal mutations, a requirement that can be rarely met in a standard clinical setting. However, the computational classification of mutations as truncal or non-truncal from a single tumor sample could provide more effective targets for individualized therapies. To this purpose, we have analyzed 12 published whole-exome sequencing datasets of matched primary-metastasis samples from 10 different cancer entities. Overall, approximately 24,000 somatic mutations were identified. The detected variants were annotated with a range of features, for instance mutation-specific features, but also sample- or gene-level features, and labeled as “shared-clonal” (proxy for truncal mutations) or “non-shared-clonal” (proxy for non-truncal mutations) based on their clonal presence or absence in the matched samples. Approximately 30% of all mutations were labeled as “shared-clonal”. However, this proportion appears to be very heterogeneous among cancer entities and patients. Exploratory data analysis also suggested that the received treatment affects the proportion of shared-clonal variants. Overall, the data exploration showed that no single variable alone could distinguish between “shared-clonal” and “non-shared-clonal” mutations. In a next step, Random Forest models with different feature sets were therefore trained to distinguish the two mutation classes. We found that the performance of these models was strongly influenced by the cancer entity used during training and that gene-level features were consistently uninformative for the models. Overall, our findings underline that tumor heterogeneity can present in multiple, biologically distinct patterns which can be taken into account for treatment decisions. The optimized model that classifies mutations as truncal or non-truncal will enable the clinical utilization of previously inaccessible information on intra- and intertumoral heterogeneity, thereby improving the efficacy of personalized immunotherapies. Citation Format: Jennifer Neumaier, Luisa Bresadola, Jonas Ibn-Salem, Ranganath Gudimella, Pablo Riesgo Ferreiro, Barbara Schrörs, Ugur Sahin. Using machine learning to tackle tumor heterogeneity [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 A032.

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 imitation

Not 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.

metaresearch head score (Codex)0.004
metaresearch head score (Gemma)0.007
Version: metacan-v3-hybrid-931329e0061cValidation status: machine_predicted_unvalidated
Candidate categoriesnone
Consensus categoriesnone
DomainCandidate signal: none · Consensus signal: none
Study designCandidate signal: Not applicable · Consensus signal: none
GenreCandidate signal: Other · Consensus signal: none
Teacher disagreement score0.005
Threshold uncertainty score0.024

Distilled classifier scores by category (both heads)

CategoryCodexGemma
Metaresearch0.0040.007
Meta-epidemiology (narrow)0.0020.001
Meta-epidemiology (broad)0.0020.002
Bibliometrics0.0020.002
Science and technology studies0.0010.001
Scholarly communication0.0020.001
Open science0.0020.002
Research integrity0.0020.002
Insufficient payload (model declined to judge)0.0030.001

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.

Opus teacher head0.197
GPT teacher head0.571
Teacher spread0.374 · how far apart the two teachers sit on this one work
Validation statusscore_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it

Classification

machine, unvalidated

Machine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.

The models applied no category: nothing in the taxonomy fit this work.
Study designNot applicable
Domainnot available
GenreOther

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".

Quick stats

Citations0
Published2025
Admission routes1
Has abstractyes

Explore more

Same venueClinical Cancer Research→Same topicRadiomics and Machine Learning in Medical Imaging→French-language works237,207→