Highlights d A machine learning (ML) workflow is designed to predict drug response in cancer patients d Deep neural networks (DNNs) surpass current ML algorithms in drug response prediction d DNNs predict drug response and survival in various large clinical cohorts d DNNs capture intricate biological interactions linked to specific drug response pathways

Myeloid-derived suppressor cells (MDSCs) represent a heterogeneous population of cells with immunosuppressive properties and might confer to worse prognosis in cancer patients. The presence of phenotypically newly described subpopulations of MDSCs and their association with the clinical outcome were investigated in non-small cell lung cancer (NSCLC) patients. The percentages and correlation between MDSCs and distinct immune cells in the peripheral blood of 110 chemotherapy-naive patients before treatment and healthy controls were investigated using flow cytometry. Two monocytic [CD14+CD15−CD11b+CD33+HLA-DR−Lin− and CD14+CD15+CD11b+CD33+HLA-DR−Lin−] and a granulocytic [CD14−CD15+CD11b+CD33+HLA-DR−Lin−] subpopulations of MDSCs were identified, expressing inducible nitric oxide synthase, and reactive oxygen species, respectively. Increased percentages of both monocytic-MDSCs' subpopulations were inversely correlated to dendritic/monocyte levels (P ≤ 0.04), while granulocytic-MDSCs were inversely correlated to CD4+ T cells (P = 0.006). Increased percentages of monocytic-MDSCs were associated with worse response to treatment (P = 0.02) and patients with normal levels of CD14+CD15+CD11b+CD33+HLA-DR−Lin− had longer overall survival and progression free-survival compared to those with high levels (P = 0.008 and P = 0.005, resp.). Multivariate analysis revealed that the increased percentages of CD14+CD15+CD11b+CD33+HLA-DR−Lin− MDSCs were independently associated with decreased progression free-survival and overall survival. The data provide evidence that increased percentages of new monocytic-MDSCs' subpopulations in advanced NSCLC patients are associated with an unfavourable clinical outcome.

Performance status and the mGPS are superior prognostic factors in advanced lung cancer. In combination, these improved survival prediction compared with either alone.

The role of the different circulating regulatory T-cells (Treg) subsets, as well as their correlation with clinical outcome of non-small cell lung cancer (NSCLC) patients is poorly understood. Peripheral blood from 156 stage III/IV chemotherapy-naive NSCLC patients and 31 healthy donors (HD) was analyzed with flow cytometry for the presence and functionality of CD4+ Treg subsets (naive, effector and terminal effector). Their frequencies were correlated with the clinical outcome. All CD4+ Treg subsets exhibited highly suppressive activity by TGF-β and IL-10 production. The percentages of naive Treg were found elevated in NSCLC patients compared to HD and were associated with poor clinical outcome, whereas the percentage of terminal effector Treg was lower compared to HD and higher levels were correlated with improved clinical response. At baseline, normal levels of naive and effector Treg were associated with longer overall survival (OS) compared to high levels, while the high frequency of the terminal effector Treg was correlated with longer Progression-Free Survival and OS. It is demonstrated, for first time, that particular CD4+ Treg subtypes are elevated in NSCLC patients and their levels are associated to the clinical outcome. The blocking of their migration to the tumor site may be an effective therapeutic strategy.

In the peripheral blood of patients with NSCLC, bevacizumab-based chemotherapy significantly reduced the levels of granulocytic MDSCs. An increase in the levels of CD15-positive monocytic MDSCs was associated with poor response to treatment and disease progression, providing evidence of their clinical relevance in patients with NSCLC.