crizotinib-naive, as demonstrated by rapid and durable responses. These findings further suggest that the activity of lorlatinib differs depending on prior exposure to crizotinib. The safety profile of lorlatinib in ROS1 patients was comparable to that previously reported in the overall ALK/ROS1 population.

period, DL1 was further assessed. 3 (50%) of 6 patients at DL1 experienced a DLT (grade 3 headaches and body pain, grade 3 neutropenia, grade 3 rash, one each). One patient was enrolled at DL -1 and did not have a DLT. The most common treatment-related adverse events (TRAEs) included pleural effusion (n¼9), diarrhea (n¼8), rash (n¼7), AST elevation (n¼6), ALT elevation (n¼6), most of which were grade 1 or 2. 4/4/1 patients had grade 1/2/3 pleural effusion, respectively. 7 (70%) patients had grade 3 TRAEs. No grade 4 or 5 toxicities were observed. Eight (80%) patients had a partial response (including 1 unconfirmed partial response) and 2 had stable disease. Median PFS was 27.2 months; median OS was not reached. The recommended phase II dose was determined as osimertinib 80 mg QD and dasatinib 70 mg QD. Pharmacokinetics (PK) analysis is being performed and will be presented. Due to slow accrual after approval of osimertinib in firstline, the trial was closed to enrollment. Conclusion: The combination of dasatinib and osimertinib demonstrated encouraging anticancer activity. Median PFS is longer than what is historically reported with osimertinib alone in first-line setting, although definitive conclusions cannot be drawn given the small sample size. The tolerability of the combination was limited by TRAEs, but they were generally manageable with dasatinib dose reductions and supportive measures.

No abstract

single nucleotide missense mutations. Method: We developed a bioinformatics pipeline for neoantigen prediction using paired normal tissue and tumor exome sequencing, RNA sequencing and HLA binding prediction. 536 lung adenocarcinoma (LUAD) and 466 lung squamous cell carcinoma (LUSC) cases were analyzed using our bioinformatics pipeline. The non-synonymous somatic SNVs and short INDELs mutations were identified to generate a list of mutation neoantigen-derived and, when possible, their corresponding wild-type epitopes. Binding affinities of the paired wild-type and mutant peptides to HLA class I were then predicted and compared. Result: On average, 8.65 (range1-158) mutant neoantigen peptides per sample were identified in 395 out of 536 (73.6%) LUAD samples. Among them, 63.7% were SNVs and 36.3% were INDELs. On average, 8.54 (range 1-504) mutant neoantigen peptides per sample were identified in 360 out of 466 LUSC samples. Among those, 67% were SNVs and 33% were INDELs. Most neoantigen peptides are private in both LUAD and LUSC. The mutant neoantigen peptides identified from INDELs were predicted to have 3.9 (p ¼ 2.42E-74) and 1.14 (p ¼ 5.44E-67) fold higher HLA class I binding affinity than wild type peptide compared to those from SNVs in LUAD and LUSC respectively. Conclusion: Tumor INDELs may be a rich source of neoantigens with a higher predicted high HLA binding affinity in lung cancers that warrant consideration in development of a personalized cancer vaccine.

Background: Tumor genomic information from a simple blood collection revealing actionable mutation can improve clinical outcome without the need for an invasive tissue biopsy. We report on the clinical utility of a cell-free DNA (cfDNA) next generation sequencing (NGS) blood test in our patients with non-small cell lung cancers (NSCLC) and S596