Abstract P2-07-07: Development of a machine learning-based classifier for Oncotype DX® category prediction in a population of lymph node positive breast carcinoma patients

INTRODUCTION: Oncotype DX® (ODX) Assay is a valuable prognostic and predictive tool in ER+, Her2- invasive breast cancer (IBC). Initially tested and validated in lymph node (LN) negative patients, the indications of this test have been expanded to include patients with limited LN-positive disease. Several prediction systems have been developed to predict the ODX Recurrence Score (RS) with substantial performance in predicting a high vs low RS score but with low performance when predicting the 3 classes that compose the standard ODX. Additionally, many of these prediction systems have not been developed and/or tested for a population of LN+ patients.

OBJECTIVES: The primary objective of this study was to evaluate the performance of several previously published ODX RS predictive systems in a population of LN+ patients. Furthermore, we developed a machine-learning based classification system to accurately predict the ODX 3 category RS for this specific population.

METHODS: We conducted a retrospective search of Stanford's pathology database for all patients with LN+ IBC diagnosed between January 2013 and December 2017 with an ODX RS available. A total of 119 patients were identified for inclusion in this cohort. Our multivariate pathologic feature-based discriminatory model aimed to classify each case as belonging to the low, intermediate or high ODX RS category. We performed model validation by the 10-fold cross validation (10F-CV) method. The model's performance was assessed by comparing simple accuracy, balanced accuracy, F1 score (harmonic average of the precision and recall) and several concordance classification metrics.

RESULTS: Of the evaluated methods, Magee equations performed well in this population of LN+ patients with the modified Magee equation 2 displaying the best accuracy (70.9%) which was surprisingly better than originally reported (55.8%, in Klein et al. Mod Path. 2013). After an initial screen of methods and tuning of the best performing model, our model achieved an overall accuracy of 78.1% on 10F-CV with a 79.1 % balanced accuracy and no two-step discordances. This corresponded to an increase of weighted Cohen's kappa of 30% versus the best performing Magee equation in this cohort and an increase of 103% versus the modified Magee 1 equation (which uses the same features as our model except for tumor grade).

DISCUSSION: Classifiers aimed at providing an alternative to Oncotype DX testing are available and perform consistently across datasets. We are currently validating our approach in a population of 1000 LN-negative patients from the MSKCC and the SEER database. Due to the substantial performance of our machine learning-based classifier based on standard reported pathologic features, our model may be considered an alternative to the ODX standard testing or a screening method for ODX testing, especially for cases with where the cost and availability of the ODX test are a concern.

Citation Format: Beca F, Yang S-R, Gruber JG, Barry-Holson K, West R, Wen HY, Allison KH. Development of a machine learning-based classifier for Oncotype DX® category prediction in a population of lymph node positive breast carcinoma patients [abstract]. In: Proceedings of the 2018 San Antonio Breast Cancer Symposium; 2018 Dec 4-8; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2019;79(4 Suppl):Abstract nr P2-07-07.

Myeloid progenitor cluster formation drives emergency and leukaemic myelopoiesis

Although many aspects of blood production are well understood, the spatial organization of myeloid differentiation in the bone marrow remains unknown. Here we use imaging to track granulocyte/macrophage progenitor (GMP) behaviour in mice during emergency and leukaemic myelopoiesis. In the steady state, we find individual GMPs scattered throughout the bone marrow. During regeneration, we observe expanding GMP patches forming defined GMP clusters, which, in turn, locally differentiate into granulocytes. The timed release of important bone marrow niche signals (SCF, IL-1β, G-CSF, TGFβ and CXCL4) and activation of an inducible Irf8 and β-catenin progenitor self-renewal network control the transient formation of regenerating GMP clusters. In leukaemia, we show that GMP clusters are constantly produced owing to persistent activation of the self-renewal network and a lack of termination cytokines that normally restore haematopoietic stem-cell quiescence. Our results uncover a previously unrecognized dynamic behaviour of GMPs in situ, which tunes emergency myelopoiesis and is hijacked in leukaemia.

Abstract 1553: Key determinants of immunotherapy success in frequently-used murine tumor models

Cancer immunotherapy uses a patient's own immune system to recognize and eliminate malignant cells. The potential advantages of this approach over conventional treatments include the systemic trafficking of immune cells to treat primary and metastatic disease, inherent antigen-specificity of adaptive immunity to minimize collateral damage, and the induction of immunological memory to prevent recurrent disease. While many acknowledge the promise of immunotherapy, its translation to the clinic has been slow and only occasionally successful. In the more than 3 decades of experience of our laboratory in developing and testing immunotherapy reagents for cancer, we observed that a given set of reagents can produce complete regressions in some but not all tumor models. We hypothesized that a major contributor to this variability was underlying differences in the immune escape strategies utilized by different tumors. Our hypothesis was supported with data from a series of immunotherapy experiments on a panel of experimental murine solid tumor models. Comprehensive immune profiles were generated for each by measuring tumor infiltrating leukocytes (TIL), immune activation, and immune suppression present in the tumor microenvironment using qRT-PCR, immunohistochemistry staining, and flow cytometry techniques. Key differences emerged amongst models in the extent of immune activation that correlated directly with TIL and immunosuppression. From these data, it appears that some tumors are more “visible” to the immune system and have active countermeasures to survive, while others use immune evasion strategies to persist in the host. To classify tumor models along this spectrum, ten immune-related genes and cell markers showing significant change across the models were selected to generate an immunogenicity score for each model: CD40, 41BBL, OX40L, CD80, CD86, CD11c, CD45, BM-2 (PMN marker), Granzyme B, and CD8. Importantly, we observed that MHC class I mouse equivalent H2-D correlated with this immunogenicity level for each tumor. Secondly, Treg and MDSC were measured in the tumor microenvironment and draining lymphoid tissues using immunohistochemical and flow cytometry techniques to determine the dominant suppressor cell population(s) present. Immunotherapy regimens were then developed to match each tumor model using immunogenicity to indicate the level of immune stimulation needed and suppressor cell component to indicate targets for reversing tumor tolerance. These regimens were used to treat established solid tumors in mice and demonstrated that two features of a tumor's immune profile, namely immunogenicity level and dominant suppressor cell component, are the key determinants of immunotherapy success in vivo. Viewed in this way, the tumor/host relationship becomes the overriding feature for determining optimal treatment for patients based upon immune profiling data obtained at the time of biopsy.

Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 103rd Annual Meeting of the American Association for Cancer Research; 2012 Mar 31-Apr 4; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2012;72(8 Suppl):Abstract nr 1553. doi:1538-7445.AM2012-1553

IL-6 controls leukemic multipotent progenitor cell fate and contributes to chronic myelogenous leukemia development

Using a mouse model recapitulating the main features of human chronic myelogenous leukemia (CML), we uncover the hierarchy of leukemic stem and progenitor cells contributing to disease pathogenesis. We refine the characterization of CML leukemic stem cells (LSCs) to the most immature long-term hematopoietic stem cells (LT-HSCs) and identify some important molecular deregulations underlying their aberrant behavior. We find that CML multipotent progenitors (MPPs) exhibit an aberrant B-lymphoid potential but are redirected toward the myeloid lineage by the action of the proinflammatory cytokine IL-6. We show that BCR/ABL activity controls Il-6 expression thereby establishing a paracrine feedback loop that sustains CML development. These results describe how proinflammatory tumor environment affects leukemic progenitor cell fate and contributes to CML pathogenesis.

Hematopoietic stem cell quiescence promotes error-prone DNA repair and mutagenesis

Most adult stem cells, including hematopoietic stem cells (HSCs), are maintained in a quiescent or resting state in vivo. Quiescence is widely considered to be an essential protective mechanism for stem cells that minimizes endogenous stress caused by cellular respiration and DNA replication. We demonstrate that HSC quiescence can also have detrimental effects. We found that HSCs have unique cell-intrinsic mechanisms ensuring their survival in response to ionizing irradiation (IR), which include enhanced prosurvival gene expression and strong activation of p53-mediated DNA damage response. We show that quiescent and proliferating HSCs are equally radioprotected but use different types of DNA repair mechanisms. We describe how nonhomologous end joining (NHEJ)-mediated DNA repair in quiescent HSCs is associated with acquisition of genomic rearrangements, which can persist in vivo and contribute to hematopoietic abnormalities. Our results demonstrate that quiescence is a double-edged sword that renders HSCs intrinsically vulnerable to mutagenesis following DNA damage.