Functional Ex-Vivo Testing Combined with Machine Learning Predicts Clinical Response to Azacitidine+Venetoclax in AML
Bibliographic record
Abstract
Background: The combination of Venetoclax + hypomethylating agents such as Azacitidine (Ven/Aza) has revolutionized the treatment of Acute Myeloid Leukemia (AML). Ven/Aza is currently the standard of care for patients with newly diagnosed AML who are not candidates for intensive chemotherapy (IC), as well as a recurrent therapeutic choice for IC-relapsed/refractory patients. Despite widespread adoption, the predictive biomarkers of Ven/Aza response/resistance available today are not sufficiently conclusive (Griffioen et al. Cancers 2022). In fact, a current hot topic in the field is whether Ven/Aza can also become a frontline therapy for fit patients as an alternative to IC. Several worldwide studies are addressing this issue, including an ongoing clinical trial named “Survival of the Fittest” exploring this concept in the US (NCT04801797). An open question is whether functional ex-vivo testing could become a powerful actionable tool that overcomes limitations presented by genetic and molecular indicators. Aim & Methods: In this study, we utilized an innovative ex-vivo functional platform of Patient Micro Avatars (PMAs) developed at OncoPrecision (García et al, Blood 2022 140 (Supplement 1): 13027-13028) to assess the predictive power for Ven/Aza treatment in a cohort of 11 AML patients. Patient-Derived Cells (PDCs) were obtained from the patients' PBMCs and Ven/Aza activity was assessed using high-throughput flow cytometry. An in-house Machine Learning (ML) tool was developed to identify the optimal variables that predict clinical outcome (Complete Remission: Yes/No). Clusters of predictive Response/Non-Response were defined by analyzing a cohort of >60 previously profiled AML patients and by leveraging the two most predictive populational features generated by the platform that better correlate with clinical outcome from >12,000 combinations of variables. Results: Our findings revealed differential ex-vivo activity of Ven/Aza in AML, with remarkably contrasting outcomes that defined clearly separate clusters of response (Figure - Left Panel). Such ML clustering led to 80% Overall Predictive Power and 100% Positive Predictive Value, meaning that all the patients for whom the platform predicted response achieved complete clinical remission. Moreover, by comparing the performance of other targeted therapies such as Gilteritinib, as well as IC regimens such as AraC-FaraA-IDA, side by side we were able to identify patients for whom Ven/Aza is the optimal regimen (e.g., Patient #169), and others for whom IC or Gilteritinib emerge as more promising therapeutic options (e.g., Patient #170) (Figure - Right Panel). Conclusions: This work presents a groundbreaking approach to predict Aza/Ven responders versus non-responders in AML. Combining the phenotypic outcomes of functional ex-vivo testing with ML tools presents the opportunity to more comprehensively predict response to cutting-edge therapies such as Ven/Aza than traditional genetic testing. This potential is particularly compelling for therapeutic schemes not directly associated with the acquisition of genetic mutations. References: 1. Griffioen MS, ... Smit L. Targeting Acute Myeloid Leukemia with Venetoclax; Biomarkers for Sensitivity and Rationale for Venetoclax-Based Combination Therapies. Cancers (Basel). 2022 Jul 15;14(14):3456. doi: 10.3390/cancers14143456. PMID: 35884517; PMCID: PMC9318140. 2. Alejandra Garcia, ... Gastón Soria; Ex-Vivo Testing Using Patient Micro Avatars (PMAs) Predicts Clinical Response in Acute Leukemias. Blood 2022; 140 (Supplement 1): 13027-13028. doi: https://doi.org/10.1182/blood-2022-164847
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) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.001 |
| Insufficient payload (model declined to judge) | 0.002 | 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 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".