Abstract 1400: Novel expression biomarkers via prediction of response to FOLFIRINOX (FFX) treatment for PDAC
Bibliographic record
Abstract
Abstract We introduce a new algorithm employing bulk RNA-seq to identify a robust set of genes associated with response, resulting in one of the first multi-gene expression biomarkers for efficacy of FFX in PDAC. Ocean Genomics TxomeAI® processed 103 pre-treatment, whole-transcriptome RNA-seq samples from the COMPASS study (obtained under agreement with University Hospital Network, Toronto) to quantify expression for every gene (GENCODE v31). Genes with 0 expression in all samples were discarded. Cases were labeled responder (RECIST CR/PR; n=23) and non-responder (SD/PD/NE; n=80) based on their response to modified FFX. 21 cases were held out for final validation. Ocean Genomics DiscoverAI™ learned a predictor for response. For each cross validation fold (100 random 80/20-splits on 82 samples), genes were selected by (1) identifying genes with statistically significant differential expression (DGE) between responders and non-responders (p < 0.01 and |Log2FC| > 0.5); (2) for each DGE gene, calculating the difference in response rate and statistical significance (Z test for proportions) between low and high expression using cutoffs obtained via the maximal chi-squared statistic; and (3) feeding the 30 genes with Z test p < 0.01 and the largest difference in response rate into permutation feature importance based on logistic regression to identify the 5 most important genes. These were used to fit a logistic regression model on the fold training data. Genes selected during the fold with the highest performance on the fold test data were used to train a model on the 82-sample training set. The average CV AUC was 0.68 (improved over 0.61 training on all genes). On the 21 held-out cases, AUC was 0.63 (compared with 0.44 if all genes were used to train a model). The genes selected by the final model were IGHG2 (Immunoglobulin Heavy Constant Gamma 2), IGKV3-20 (Immunoglobulin Kappa Variable 3-20), IGLL5 (Immunoglobulin Lambda Like Polypeptide 5), WASH8P (pseudogene associated with Wiskott-Aldrich syndrome), and HBB (Hemoglobin Subunit Beta). Three of these genes are related to immunoglobulin, and there is literature to support that immune-complex-bound proteins are predictive of response to chemotherapy. All genes in the final model were among the top-7 most frequently selected genes across folds (selected between 23 and 48 times out of the 100 folds). hENT1, a previous biomarker of GA effectiveness, also shows separation of the survival curves on this data set, but less so than several of the identified genes. While additional biological validation and additional computational variations of the study design are required to confirm the genes and predictor, this analysis resulted in a robust candidate set of 5 genes from a large PDAC, FFX-treated cohort that individually are statistically significantly associated with response, and that together are more predictive of response to FFX in this data than all-gene models. Citation Format: Hossein Asghari, Ehsan Haghshenas, Roby Thomas, Eric Schultz, Rob Patro, Stan Skrzypczak, Carl Kingsford. Novel expression biomarkers via prediction of response to FOLFIRINOX (FFX) treatment for PDAC [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 1400.
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 | 0.001 | 0.001 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.000 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.002 | 0.001 |
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.
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".