Joint Analysis of Multiple Algorithms and Performance Measures

There has been an increasing interest in the development of new methods using Pareto optimality to deal with multi-objective criteria (for example, accuracy and time complexity). Once one has developed an approach to a problem of interest, the problem is then how to compare it with the state of art. In machine learning, algorithms are typically evaluated by comparing their performance on different data sets by means of statistical tests. Standard tests used for this purpose are able to consider jointly neither performance measures nor multiple competitors at once. The aim of this paper is to resolve these issues by developing statistical procedures that are able to account for multiple competing measures at the same time and to compare multiple algorithms altogether. In particular, we develop two tests: a frequentist procedure based on the generalized likelihood-ratio test and a Bayesian procedure based on a multinomial-Dirichlet conjugate model. We further extend them by discovering conditional independences among measures to reduce the number of parameters of such models, as usually the number of studied cases is very reduced in such comparisons. Data from a comparison among general purpose classifiers is used to show a practical application of our tests.

Activation of STING-Dependent Innate Immune Signaling By S-Phase-Specific DNA Damage in Breast Cancer

BACKGROUND: Previously we identified a DNA damage response-deficient (DDRD) molecular subtype within breast cancer. A 44-gene assay identifying this subtype was validated as predicting benefit from DNA-damaging chemotherapy. This subtype was defined by interferon signaling. In this study, we address the mechanism of this immune response and its possible clinical significance.

METHODS: We used immunohistochemistry (IHC) to characterize immune infiltration in 184 breast cancer samples, of which 65 were within the DDRD subtype. Isogenic cell lines, which represent DDRD-positive and -negative, were used to study the effects of chemokine release on peripheral blood mononuclear cell (PBMC) migration and the mechanism of immune signaling activation. Finally, we studied the association between the DDRD subtype and expression of the immune-checkpoint protein PD-L1 as detected by IHC. All statistical tests were two-sided.

RESULTS: We found that DDRD breast tumors were associated with CD4+ and CD8+ lymphocytic infiltration (Fisher's exact test P < .001) and that DDRD cells expressed the chemokines CXCL10 and CCL5 3.5- to 11.9-fold more than DNA damage response-proficient cells (P < .01). Conditioned medium from DDRD cells statistically significantly attracted PBMCs when compared with medium from DNA damage response-proficient cells (P < .05), and this was dependent on CXCL10 and CCL5. DDRD cells demonstrated increased cytosolic DNA and constitutive activation of the viral response cGAS/STING/TBK1/IRF3 pathway. Importantly, this pathway was activated in a cell cycle-specific manner. Finally, we demonstrated that S-phase DNA damage activated expression of PD-L1 in a STING-dependent manner.

CONCLUSIONS: We propose a novel mechanism of immune infiltration in DDRD tumors, independent of neoantigen production. Activation of this pathway and associated PD-L1 expression may explain the paradoxical lack of T-cell-mediated cytotoxicity observed in DDRD tumors. We provide a rationale for exploration of DDRD in the stratification of patients for immune checkpoint-based therapies.