Abstract We have been comprehensively screening for “vulnerabilities” that have been acquired during the multi-step pathogenesis of lung cancer cells but are not present in normal lung epithelial cells to identify genetic and chemical perturbations that will selectively kill lung cancer. We think many of these have occurred to allow the lung cancers to undergo/tolerate “oncogene addiction.” We tested a sub-panel of 12-15 non-small cell lung cancer (NSCLC) lines that covers the known molecular spectra of lung cancer with genome wide siRNA and large scale chemical library (~250,000 compounds) and natural products in vitro screens to identify “hits” that will kill (suppress the growth of) lung cancer cells but not normal human bronchial epithelial cells and that also only kill a subset of lung cancer cells providing two types of specificity. “Hits” from these broad screens are then tested (including detailed drug concentration curves) across a large panel of lung cancer lines (~100) representing a variety of lung cancer histologic and molecular oncogenotypes. Other versions of these screens include the intensive use of “mini-libraries” each containing 50 – 150 gene targets by siRNAs or shRNAs, or ~200 defined drugs to explore pathways in detail in tests of over 70 NSCLCs. Examples include: nuclear receptors and their co-regulators (120 genes); cancer stem cell pathways (50 genes); chromatin remodelers (75 genes) and identified lung cancer mutated driver oncogenes (175 genes). In addition to the in vitro tests, we have developed in vivo (xenograft) tests where shRNA mini-libraries are introduced into tumor cells at high representation which are grown as xenografts, analyzed by NexGen sequencing and shRNAs identified that drop out or are retained in xenografts compared to in vitro grown cells to identify vulnerabilities that are only detected in the in vivo situation. All of the data are then related to the large legacy molecular datasets associated with the lung cancer lines (including whole exome sequence analyses and genome wide mRNA, copy number variation, methylation, miR expression data and proteomics data). In addition, detailed chemical and pharmacokinetic analyses for favorable drug properties and subsequent chemical modifications also occur for the chemical compounds to progress those towards potential clinical studies. The results of all of these analyses have identified ~300 new chemical compounds and ~300 genetic hits all of which show selectivity for lung cancer over normal lung cells and selectivity for subtypes of lung cancer. The chemical and genes hits are being compared to the tumor molecular information and integrated in turn through a “connectivity map” type of approach – to identify drugs and gene hits involving the same pathways. The molecular correlates of the tumor lines are related to similar molecular changes in patient derived xenografts and patient tumor specimens to provide a connection of the molecular subtype-selective vulnerabilities (“enrollment biomarkers”) between the preclinical response phenotypes and patient tumor specimens. From these data we find lung cancers can be classified into groups (“clades”) that represent functional vulnerabilities to the gene and chemical compound hits and these in turn can be related to molecular abnormalities in tumors. One example of this is our detailed analyses a matched lung adenocarcinoma/normal lung epithelial cell model derived from the same patient which identified three distinct target/response-indicator pairings that are represented a significant frequencies (6-16%) in the lung adenocarcinoma population (Kim et al. Cell 155:552, 2013). These include three totally novel lung cancer selective targeted therapies: NLRP3 mutation/inflammasome activation-dependent FLIP addiction; co-occurring KRAS and LKB1 mutation-driven COPI addiction; and selective sensitivity to a synthetic indolotriazine that is specified by a seven-gene expression signature. Our panel of “hits” provide the opportunity to identify all potential therapeutic targets for lung cancer, while the molecular correlates will allow “personalization” of these new therapies going forward in preclinical and clinical translation. (Supported by NCI SPORE P50CA70907, NCI CTD2N, CPRIT, UTSW CCSG P30CA142543) Citation Format: John D. Minna, Adi Gazdar, Alexander Augustyn, Rebecca Britt, Ryan Carstens, Patrick Dospoy, Boning Gao, Luc Girard, Suzie Hight, Kenneth Huffman, Jill Larsen, Michael Peyton, Chunli Shao, David Mangelsdorf, Rolf Brekken, Ralph Deberardinis, Pei-Hsuan Chen, Carmen Behrens, Lauren Byers, John Heymach, Jack Roth, Ignacio Wistuba, Yang Xie, Caleb Davis, David Wheeler, Richard Gibbs, Edward Marcotte, Joseph Ready, Deepak Nijhawan, Noelle Williams, Steven McKnight, Bruce Posner, John MacMillan, Michael Roth, Michael White. Developing a new functional classification of lung cancer based on tumor acquired vulnerabilities. [abstract]. In: Proceedings of the AACR-IASLC Joint Conference on Molecular Origins of Lung Cancer; 2014 Jan 6-9; San Diego, CA. Philadelphia (PA): AACR; Clin Cancer Res 2014;20(2Suppl):Abstract nr IA21.