AU2017236791B2 - ERRC1 and other markers for stratification of non-small cell lung cancer patients - Google Patents

Abstract

Proteomics data are employed to predict treatment outcome for cisplatin/pemetrexed therapy of NSCLC. Notably, ERRC1 can be used as a binary marker where protein presence or absence is measured by mass spectroscopy, and additional proteins may serve as further predictive markers, particularly for prediction of PFS.

Description

ERCCl AND OTHER MARKERS FOR STRATIFICATION OF NON-SMALL CELL

LUNG CANCER PATIENTS

[0001] This application claims priority to US provisional application serial number

62/311,368, filed March 21, 2016, and also claims priority to U.S. provisional application serial number 62/337,209, filed May 16, 2016; both provisional applications are incorporated herein by reference.

Field of the Invention

[0002] The field of the invention is systems and methods of association of selected markers with clinical outcome, especially as it relates to non-small cell lung cancer and ERCCl status and recurrence free survival/overall survival.

Background of the Invention

[0003] The background description includes information that may be useful in understanding the present invention. It is not an admission that any of the information provided herein is prior art or relevant to the presently claimed invention, or that any publication specifically or implicitly referenced is prior art.

[0004] All publications and patent applications herein are incorporated by reference to the same extent as if each individual publication or patent application were specifically and individually indicated to be incorporated by reference. Where a definition or use of a term in an incorporated reference is inconsistent or contrary to the definition of that term provided herein, the definition of that term provided herein applies and the definition of that term in the reference does not apply.

[0005] ERCCl, the excision repair 1 endonuclease non-catalytic subunit, functions in the nucleotide excision repair pathway and is required for the repair of DNA lesions such as those induced by UV light or formed by electrophilic compounds, including cisplatin. The ERCCl protein forms a heterodimer with the ERCC4 endonuclease, and the heterodimeric endonuclease catalyzes the 5 '-incision in the process of excising the DNA lesion. The heterodimeric endonuclease is also involved in recombination DNA repair and in the repair of inter-strand crosslinks. Mechanistically, ERCCl appears to play an at least somewhat significant role in various cancers. For example, deficient expression of the DNA repair enzyme ERCCl was reported to associate with colon cancer (Genome Integr. 2012; 3: 3; Scientific Reports 2014, 4: 4313), while transcription of the ERRC1 gene was shown to be reduced in a significant fraction of gliomas (Int J Cancer 2010 Apr 15;126(8): 1944-54).

[0006] While at least from an observational perspective the ERCCl status appeared to be a promising candidate, no currently known system has provided statistically significant models that can predict recurrence free survival (PFS)/overall survival (OS) that employ ERCCl as an indicator, especially for cisplatin-based treatments. Indeed, it was previously hypothesized that low or absent tumor expression of the ERCCl protein could predict improved survival in platinum-treated NSCLC patients. However, a phase 3 of the Tailored Postsurgical Therapy in Early-Stage NSCLC (TASTE) trial was canceled because ERCCl assessment performed by immunohistochemistry has proved unreliable.

[0007] Thus, even though numerous methods of ERCCl tests are known in the art, all or almost all of them suffer from various disadvantages. Consequently, there remains a need for improved systems and methods to predict recurrence free survival/overall survival in cancer, and especially NSCLC.

Summary of The Invention

[0008] The inventive subject matter is directed to compositions and methods of predicting recurrence free survival and/or overall survival in cancer upon treatment, especially where the cancer is NSCLC and the treatment is cisplatin/pemetrexed.

[0009] In one aspect of the inventive subject matter, the inventors contemplate a method of predicting overall survival and/or progression free survival for a patient with non-small cell lung cancer that is subject to treatment with cisplatin/pemetrexed. Such method typically includes a step of obtaining quantitative mass spectroscopic data for ERCCl in a patient sample and another step of classifying the quantitative data as 'not detectable' when the quantitative data is below a threshold value, and as 'detectable' when the quantitative data is above a threshold value. In still another step, a patient record is updated or generated to denote that the patient has an improved (relative to patients having 'detectable' levels of ERCCl) overall survival and/or progression free survival after treatment with

cisplatin/pemetrexed when the quantitative data are 'not detectable'.

[0010] In further contemplated aspects, additional quantitative mass spectroscopic data may be obtained for E-cadherin, HER2, TITF1, MSLN, KRT7, FR-alpha, HER3, FPGS, and/or ROS l in the patient sample, typically to predict progression free survival. In general, progression free survival is increased when expression of the E-cadherin, HER2, TITFl , MSLN, KRT7, FR-alpha, HER3, FPGS, and/or ROS l is increased (e.g., relative to average across a large number patients with NSCLC). Most typically, the patient sample is fresh biopsy material, a frozen biopsy sample, or a formalin fixed paraffin embedded sample.

[0011] The quantitative mass spectroscopic data is typically, but not necessarily, obtained from selected reaction monitoring mass spectroscopy (SRM-MS), and/or the threshold value is between two and five times of a standard deviation of background signal (e.g., equal or less than 1.0 fmol, or 0.5 fmol, or 0.1 fmol). It is further contemplated that the patient record is an electronic record, which may be generated or updated to include a treatment

recommendation to administer cisplatin/pemetrexed.

[0012] Therefore, and viewed from a different perspective, the inventors also contemplate use of quantitative mass spectroscopic data for ERCC1 in a patient sample to predict at least one of overall survival and progression free survival of a patient with non-small cell lung cancer subject to treatment with cisplatin/pemetrexed. As noted above, the quantitative data is preferably classified as 'not detectable' when the quantitative data is below a threshold value, and as 'detectable' when the quantitative data is above a threshold value. The patient is then determined as having an improved overall survival and/or progression free survival after treatment with cisplatin/pemetrexed when the quantitative data are 'not detectable' .

[0013] Contemplated uses may further comprise use of expression of E-cadherin, HER2, TITFl , MSLN, KRT7, FR-alpha, HER3, FPGS, and/or ROS l in the patient sample to predict progression free survival of the patient, where most typically the progression free survival is increased when expression of the at least one of E-cadherin, HER2, TITFl , MSLN, KRT7, FR-alpha, HER3, FPGS, and ROS l is increased (e.g., relative to average across a large number patients with NSCLC).

[0014] Various objects, features, aspects, and advantages of the inventive subject matter will become more apparent from the following detailed description of preferred embodiments, along with the accompanying drawing figures.

Brief Description of The Drawings

[0015] Figure 1 shows an exemplary flow chart for data analysis as performed herein. [0016] Figures 2A and 2B are exemplary Kaplan Meier curves of trial patients dichotomized by ERCC1 proteomics levels for overall survival (2A) and recurrence free survival (2B).

[0017] Figures 3A and 3B are exemplary plots showing the lack of concordance for ERCC1 levels as determined by IHC and MS proteomics, with Figure 3A depicting a waterfall plot for continuous measurement and Figure 2B depicting a mosaic plot for binary measurement.

[0018] Figure 4 is an exemplary heatmap of pairwise correlation between 38 proteomic markers across 146 patients in the trial.

[0019] Figures 5A and 5B are exemplary graphs from unsupervised clustering of patients based on measured markers. Figure 5A is a heatmap of scaled marker levels, while Figure 5B shows Kaplan Meier curves for recurrence free survival stratified by patient subtypes.

[0020] Figure 6 is an exemplary Kaplan Meier curve of trial patients dichotomized based on predicted relative risk using binary ERCC1 status and KRT7.

Detailed Description

[0021] The inventive subject matter is drawn to systems and methods for improved statistical and biochemical analyses for various cancers, and especially predictive analysis for treatment of non-small cell lung cancer (NSCLC) with a combination of cisplatin and pemetrexed. In especially preferred aspects, the inventors have unexpectedly discovered that when ERCC1 protein levels are quantitatively determined using methods other than immunohistochemistry (and especially quantitative mass spectroscopic proteomic analysis of a tumor sample), the so obtained ERCC1 results had significant statistical power to predict overall survival and/or progression free survival, particularly in combination with expression levels of selected other markers that were discovered. Such finding was particularly unexpected as ERCC1 levels as determined by immunohistochemistry were previously deemed unreliable (see / Clin Oncol 2014, 32: 1256-1261).

[0022] In addition, further quantitative proteomics analysis revealed that additional markers could be taken into account to predict treatment outcome, and especially PFS at statistically significant power. The terms "recurrence free survival" (RFS) and "progression free survival" (PFS) are used interchangeably herein. Notably, and as is shown in more detail below, the markers could be grouped into three classes, with one class being predictive for increased PFS as evidenced in a Kaplan Meier curves.

[0023] Therefore, and based on the results as also shown in more detail below, the inventors contemplate a method of predicting overall survival and/or progression free survival for a patient with non-small cell lung cancer where the patient is to be treated with a combination of cisplatin and pemetrexed. Such prediction can be used to evaluate a patient's eligibility for treatment with one or more drugs interfering with DNA repair, or to stratify a patient group into patients that are susceptible to treatment with one or more drugs interfering with DNA replication and repair. For example, suitable drugs that interfere with DNA replication and repair include antineoplastic drugs that inhibit one or more of base excision repair, nucleotide excision repair, DNA polymerases, homologous recombination repair, double strand break repair, and PARP. Viewed from a different perspective, quantitative mass spectroscopic data for ERCC1 in a patient sample can be used to predict overall survival and/or progression free survival of a patient with NSCLC subject to treatment with one or more drugs that interfere with DNA replication and repair (and especially cisplatin/pemetrexed)

[0024] With respect to contemplated samples it is generally preferred that the samples will originate from a patient that is diagnosed with NSCLC, and all such tumor samples are deemed suitable for use herein, including fresh biopsy samples, frozen biopsy samples, and formalin fixed paraffin embedded samples. Most typically, contemplated samples will be suitable for proteomics analysis, and particularly quantitative proteomics analysis. However, in alternative aspects, it is contemplated that the samples need not be limited to NSCLC, but may be samples from numerous other tumors, and especially tumors that are sensitive to treatment with one or more drugs that interfere with DNA replication and repair, especially cisplatin and pemetrexed. Thus, suitable tumors include those obtained from a patient diagnosed with testicular cancer, ovarian cancer, breast cancer, bladder cancer, head and neck cancer, cervical cancer, lung cancer, mesothelioma, esophageal cancer, brain tumors, and neuroblastoma.

[0025] It should further be appreciated that contemplated samples will be processed in one or more workflow to so obtain protein specific quantitative expression results. Depending on the particular type of sample, it should be recognized that the manner of obtaining quantitative results may vary, and that all manners of protein quantification are deemed suitable for use herein. For example, preferred analytic methods include various mass spectroscopic methods that may use a cell extract or even an FFPE sample as starting material. Where desired, the sample may be enriched in ERCCl and/or one more proteins of interest, or may be directly used. Among other methods, various selected reaction monitoring methods (e.g., consecutive reaction monitoring, multiple reaction monitoring, parallel reaction monitoring) are preferred. In still further contemplated quantitative methods, amounts of ERCCl and other proteins may be quantified via relative quantification methods such as isotope-coded affinity tags (ICAT), isobaric labeling (tandem mass tags (TMT) and isobaric tags for relative and absolute quantification (iTRAQ)), label-free quantification metal-coded tags (MeCAT), N-terminal labeling, stable isotope labeling with amino acids in cell culture (SILAC), and terminal amine isotopic labeling of substrates (TAILS). Still further, quantification may also include indirect methods, including isotope and/or fluorophor labeling using protein specific ligands.

[0026] Therefore, it should be appreciated that the particular type of data obtained may vary considerably. However, it is generally preferred that the quantitative protein raw data will be transformed into absolute protein values. For example, absolute protein values will typically be normalized to a specific parameter. For example, the absolute quantitative data may be expressed as absolute weight per unit sample measured (e.g., expressed in picogram per sample), as relative expression to a fixed metric (e.g., fmol protein/microgram total protein, or fmol/1000 cells), or as relative abundance compared to reference protein (e.g., % of actin) or as relative expression level as compared to an average or median value representative for the protein in tumors of patients diagnosed with a particular cancer (e.g., n-fold over- or under-expression, linear or log-based).

[0027] With respect to ERCCl data it is then preferred that the quantitative data are classified into a binary schema to provide a qualitative representation of the data. For example, such representation may be classified as 'not detectable' (or absent, or zero, or N/D, etc.) when the quantitative data is below a predetermined threshold value, and as 'detectable' (or present, or one, or positive, etc.) when the quantitative data is above a predetermined threshold value. While such transformation appears to entirely negate the benefit of a highly selective and specific (typically mass spectroscopic) analysis, the inventors' results have shown that the binary transformation unexpectedly affords statistically significant prediction results that were not achieved with previously known quantitative IHC methods.

[0028] Most typically, the threshold value for a classification as being 'not detectable' will be at or near the detection limit for a particular quantification method and will as such at least in part be dependent on the specific analytic method employed. Thus, in one aspect of the inventive subject matter, the threshold value may be two times, or three times, or four times, or five times the value of a standard deviation of the background signal in the quantitative measurement. For example, for most mass spectroscopic methods, suitable threshold values include O. lfmol, or 0.5 fmol, or 1.0 fmol, or 5 fmol, or 10 fmol (e.g., per cell), and in some cases even higher. Therefore, all quantitative ERCCl values above the threshold value will be classified as 'detected' . Viewed from a different perspective, ERCCl values may be classified 'detected' where the measured quantity of ERCCl exceeds two times, or three times, or four times, or five times the value of a standard deviation of the background signal in the quantitative measurement. For example, absolute values for ERCCl of 100 fmol or 250 fmol, or 2,000 fmol may be classified 'detected' .

[0029] With preset to the quantification of additional marker proteins (and especially those as discussed below), quantification need not be converted into a binary schema. Most typically, but not necessarily, quantification of additional marker proteins will be performed using the same methods as the quantification of ERCCl, however, the data will typically be expressed as up-regulated (over-expressed) or down-regulated (under-expressed), typically relative to a normal expression level in a non-tumor cell (or in some cases relative to the mean or median value of expression of tumor tissues, regardless of their sensitivity towards the drug or drugs that interfere with DNA replication and repair.

[0030] As discussed in more detail below, association of the results with increased overall survival and/or progression free survival is then performed on the basis of the classification of the quantitative results for ERCCl, and where desired, the expression levels of selected additional markers. Surprisingly, improved overall survival and progression free survival after treatment with cisplatin/pemetrexed was found when the quantitative data for ERCCl was 'not detectable' . Accordingly, such association can be recorded into a new or existing patient record. Improved overall survival is relative to overall survival results for patients having 'detectable' levels of ERCCl, as is seen in studies performed herein and as evidenced by Kaplan Meier curves. Similarly, where the additional marker proteins (and especially E- cadherin, HER2, TITF1 , MSLN, KRT7, FR-alpha, HER3, FPGS, and/or ROS 1) are measured, progression free survival is substantially improved in patients that exhibit over- expression of the additional marker proteins as is seen in studies performed herein and as evidenced by Kaplan Meier curves. Examples

[0031] Clinical and biomarker data from 150 patients from the TASTE Trial (IFCT-0801 TASTE) was received in excel format, saved in .txt format, and read into R for data preprocessing and statistical analysis. Consort diagram detailing patients available for primary and secondary analyses of association of ERCCl with recurrence free survival (PFS) and overall survival (OS), as well as exploratory analyses evaluating correlations between various biomarkers is shown in Figure 1 below. ERCCl proteomic data were obtained from FFPE biopsy specimens subjected to laser micro-dissection to obtain tumor tissue and subsequent selected reaction monitoring mass spectroscopy.

[0032] Input proteomic biomarker data was converted to numeric variables, where "ND", which denotes non-detectable levels of a given biomarker, is taken as 0, and "NR", which indicates missing data, is replaced with NA. Survival endpoints were defined as per the STEEP system. Overall survival time was computed as the time between date of inclusion and date of death, while recurrence free survival (RFS) time was calculated as the time between date of inclusion and date of recurrence or date of death (whichever was earlier). Patients who did not experience an event (recurrence or death) were censored at their last known follow-up.

[0033] Primary outcome analysis was performed as follows: ERCCl proteomic levels was evaluated as both a continuous variable as well as a binary variable (negative, i.e. =0 vs. positive, i.e. >0). A univariate Cox proportional hazard model was used to assess whether continuous and/or binary ERCCl levels were associated with OS and/or RFS, with significance assessed using the likelihood ratio test. No adjustments for multiplicities were employed, and Table 1 below shows the results from the univariate Cox analysis (* denotes a hazard ratio associated with 1 standard deviation increase of ERCC levels).

[0034] As a continuous variable, ERCCl proteomic levels did not significantly associate with OS or PFS. However, the dichotomized ERCCl groups showed a significant association with OS in the univariate Cox analysis, where ERCCl -positive patients had an increased risk of death when compared to ERCCl -negative patients (HR=1.3, p =0.02). A similar trend for association was observed between binary ERCCl groups and PFS. [0035] Kaplan Meier survival curves of dichotomized ERCCl groups (by proteomics levels) are shown in Figure 2 where panel (A) shows overall survival, and panel (B) shows recurrence free survival.

Table 1 - Univariate Cox Proportional Hazard Modeling of Survival on ERCCl Proteomic Levels

[0036] Secondary outcome analysis was performed as follows: Multivariate Cox proportional hazard models adjusting for stage, smoking status, age and gender was used to evaluate whether continuous and/or binary ERCCl levels were associated with OS and/or PFS independent of these clinical covariates. Significance was assessed as the likelihood ratio p value associated with the ERCCl term when it was added to a model containing the clinical covariates.

[0037] To mitigate model convergence issues, the inventors first evaluated associations between continuous and/or binary ERCCl levels and clinical covariates. For each clinical variable that associates with ERCCl levels, the inventors used bivariate Cox models to adjust for one clinical covariate at a time to assess whether ERCCl levels remains independently associated with OS/RFS. Table 2 (* denotes hazard ratio associated with 1 standard deviation increase of ERCC levels;† denotes hazard ratio associated with 1 year increase in age) shows the results from multivariate Cox overall survival analysis. Binary ERCCl groups remain independently associated with OS upon adjusting for stage, smoking, age and gender. Table 3 (* denotes hazard ratio associated with 1 standard deviation increase of ERCC levels;† denotes hazard ratio associated with 1 year increase in age) shows the results from multivariate Cox recurrence free survival analysis. Neither continuous nor binary ERCCl levels are associated with RFS.

Table 2

Current Smoking Ref Current Smoking Ref

Never Smoked Never Smoked -

Quit Smoking 3.11 (0.29 - 33.4) Quit Smoking 2.70 (0.24 - 29.9)

Gender Gender

F Ref F Ref

M 0.66 (0.15-2.83) M 0.67 (0.16-2.81)

Age 1.04 (0.95 - 1.12)† Age 1.04 (0.96 - 1.14)†

Table 3

[0038] Table 4 below summarizes the distribution of continuous and/or binary proteomic ERCCl levels stratified by stage, smoking status and gender. No significant associations were observed between continuous ERCCl levels with these clinical variables. There was no difference in the proportion of proteomic ERCCl -positive patients between stages or gender, although there appeared to be a higher proportion of proteomic ERCCl -positive patients among patients who quit smoking. ERCCl was not associated with age, when considered either as a continuous variable (Rp= -0.06, p =0.51) or a continuous variable (Wilcoxon rank sum test p=0.41).

Table 4

IIB (n=42) 67 (0 - 137) 36 (86%)

IIIA/IV (n=27) 62 (0 - 100) 19 (70%)

Smoking Kruskal 0.17 Fisher Exact 0.02

Wallis test

Current Smoking 43 (0 - 115) 9 (56%)

(n=16)

Never Smoked (n=9) 65(0 - 88) 5 (55%)

Quit Smoking (n=97) 66 (0 - 139) 80 (82%)

Gender Wilcoxon 0.42 Fisher Exact 1.00 rank sum test

F (n=44) 62.5 (0 - 128) 34 (77%)

M (n=78) 65.5 (0-139) 60 (77%)

[0039] Based on the above analyses, the inventors employed a bivariate model of overall survival as a function of binary proteomic ERCCl levels and smoking status, and observed that binary ERCCl levels remains significantly associated with OS (likelihood ratio p =0.03) after adjusting for smoking status.

[0040] Exploratory Analysis - Correlation Between Biomarkers: 146 patients with biomarker data were available for analysis as shown in Figure 1.

[0041] Association between proteomic and IHC (immunohistochemistry) ERCCl levels: The inventors compared continuous and binary proteomic ERCCl levels with IHC ERCCl status. First, the inventors evaluated whether there is a significant difference in proteomic ERCCl levels between patients within the three IHC-defined ERCCl subsets (Positive, Negative, Indeterminate) using ANOVA. As well, the inventors assessed whether there is a significant difference in the distribution of proteomic ERCCl -positive patients within the three IHC- defined subsets using a chi-square test.

[0042] The inventors did not observe any significant associations between proteomic ERCCl levels, either as a continuous variable or a binary variable, and ERCCl IHC status. Figure 3A shows a waterfall plot of ERCCl protein levels within each IHC-defined ERCCl subset; while Figure 3B shows a mosaic plot of patient with detectable proteomic ERCCl levels by ERCCl IHC status. Altogether, the analyses suggest that concordance between proteomic ERCCl levels and ERCCl IHC status is low, where 12 of 36 IHC ERCCl -positive patients has no detectable proteomic ERCC levels and 70 of 88 IHC ERCCl -negative patients has detectable ERCCl levels by proteomics.

[0043] Correlation between proteomics biomarkers: 49 further biomarkers were assessed by proteomics within this TASTE trial cohort. 4 biomarkers (FGFR2, FGFR3, HER4, MDM2) were not detectable in all 146 patients, while another 7 markers (KRAS, PTEN, HGF, MRPl, MCL1, DHFR, TLE3) had missing values in over 20% of patients. These biomarkers were filtered out, resulting in 38 biomarkers available for correlation analysis. Pairwise Pearson correlation was computed between each biomarker pair across patients, and unsupervised clustering (average linkage) was performed. Moreover, the squared Euclidean distance between each patient pair based on their biomarker data, and clustering of patients based on the Ward minimum variance method was conducted. Association between patient subsets from this unsupervised analysis and survival (OS and RFS) was assessed using a Cox proportional hazard model. Archived tumor tissues were microdissected and solubilized. In each liquefied sample, 40 protein biomarkers including ERCC1 were quantitated with a multiplexed mass spectrometry assay.

[0044] Figure 4 shows the heatmap of pairwise correlations between the 38 biomarkers across 146 patients. Overall, the pairwise correlations between markers appear modest (median = -0.009, range = -0.33 to 1). Unsupervised clustering yields 3 major subsets.

Notably, ERCC1 co-clusters (in biomarker cluster C3) with genes associated with DNA replication and repair such as hENTl, XRCC1, TYMS, TOPOl, and TOP02A.

[0045] When patients were clustered based on their biomarker profile, three distinct patient subsets were observed as can be seen from Figure 5A. These patient subsets have distinct RFS outcomes, where patients with higher protein levels of cluster CI biomarkers appear to have better RFS than those with higher levels of C3 biomarkers as is shown in Figure 5B. No significant differences in OS between patient subsets were observed. Figure 5A is a heatmap of scaled (to mean 0 and standard deviation 1) biomarker levels, and patients are arranged along the columns; and the biomarkers are along the rows. Patient subtypes are I (yellow), II (orange), III (red) are highlighted using boxes. Figure 5B depicts Kaplan Meier survival curves showing RFS stratified by patient subtypes, with hazard ratio and Ward p values displayed in the legend.

[0046] Exploratory Analysis - Additional biomarkers associated with survival outcome: In addition to ERCC1, the inventors evaluated whether other proteomic biomarkers, as a continuous or binary (detectable vs. not), is associated with OS and/or RFS using a univariate Cox proportional hazard model. The analysis is performed using the 122 CDDP-pemetrexed treated patients with available outcome and biomarker data. [0047] Initially, the 38 biomarkers evaluated in the correlation analysis were considered. However, upon further review of the data, an additional filtering criterion of having non-zero levels in at least 20% of samples was implemented, leaving a set of 22 biomarkers available for this analysis. Of these 22 biomarkers, four (GART, KRT7, Vimentin and TYMP_DGP) has detectable levels in all 122 patients considered in this analysis and were thus evaluated only as continuous biomarkers. Altogether, 40 biomarker variables (22 continuous, 18 binary) were each individually tested, and the Benjamini-Hochberg False Discovery Rate (BH FDR) correction was used to adjust for multiple hypothesis testing.

[0048] Table 5 below shows results from the univariate Cox analysis of OS as a function of biomarker levels. Six of the biomarker variables, including binary ERCC1 levels, tested showed significant univariate associations with OS prior to multiple testing adjustments. However, only binary E-cadherin levels retained significance following BH FDR correction. Review of the binary E-cadherin data revealed that only four patients have non-detectable E- cadherin levels. Two of these patients died. Although this finding is of statistical significance (BH FDR corrected p < 0.05), the E-cadherin negative subset size is so small that this result needs to be interpreted with caution. Table 6 below shows results from the univariate Cox analysis of RFS as a function of biomarker levels. Although six of the variables tested shows univariate association with RFS, significance was not retained following multiple testing correction.

Table 5

ERCC1 1.30 (0.65-2.62) 0.45 0.82

KRT5_binary 0.56 (0.12-2.68) 0.49 0.85

TUBB3_binary — 0.52 0.87

XRCC1 1.16 (0.72-1.86) 0.56 0.90

SPARC 0.82 (0.38-1.75) 0.59 0.91

GART 0.85 (0.44-1.65) 0.63 0.93

Fralpha_binary 1.21 (0.35-4.17) 0.77 0.93 hENTl_binary 0.66 (0.08-5.32) 0.71 0.93

HER2 1.12 (0.66-1.92) 0.68 0.93

IDOl _.binary 0.84 (0.24-3) 0.79 0.93

TITF1 0.9 (0.46-1.76) 0.76 0.93

TOP02A_binary 1.23 (0.3-4.97) 0.77 0.93

TUBB3 1.15 (0.52-2.52) 0.74 0.93

Fralpha 0.93 (0.45-1.94) 0.86 0.95

TOPOl 0.93 (0.45-1.9) 0.84 0.95

MSLN_binary 0.91 (0.25-3.25) 0.88 0.95

Vimentin 1.05 (0.49-2.22) 0.91 0.95

MGMT 1.01 (0.62-1.67) 0.96 0.98

MSLN 1.00 (0.43-2.31) 1.00 1.00

Table 6

TUBB3_Binary 0.67 (0.09-5.03) 0.72 0.89

GART 1.06 (0.75-1.49) 0.75 0.90

HER2 0.94 (0.65-1.37) 0.76 0.90

Fralpha 0.97 (0.62-1.5) 0.88 0.98

XRCCl_Binary 1.16 (0.16-8.6) 0.88 0.98

HER2_Binary 0.97 (0.43-2.17) 0.93 0.98

TOP01_Binary 0.93 (0.22-3.97) 0.92 0.98

TOPOl 0.99 (0.67-1.47) 0.97 0.99

Fralpha_Binary 1.00 (0.46-2.17) 1.00 1.00

[0049] Exploratory Analysis - Multi-gene predictor of survival outcomes: Elastic net for the Cox model (implemented in glmnet package in R) with cross-validation was used to construct multi-gene predictors of survival outcomes. 22 continuous biomarker variables, along with binary ERCC1 status, were considered. This analysis was restricted to a subset of 80 patients with complete biomarker data. As ERCC1 is a primary variable of interest, the shrinkage for the binary ERCC1 status was set to 0, such that it will always be included in the model. Model yielding the minimum (3-fold) cross-validation partial-likelihood is selected.

[0050] An elastic net model containing KRT7 with ERCC1 binary status appeared to give optimal prediction of the relative risk of recurrence. Using this model, patients with predicted relative risk < 1 have significantly better RFS than those with predicted relative risk > 1 (likelihood ratio p = 0.02). The Kaplan Meier survival plots of these patient subsets are shown in Figure 6 illustrating the curves of patients dichotomized based on predicted relative risk using an elastic net model of binary ERCC1 status and KRT7.

[0051] Protein biomarkers of response to pemetrexed were also quantified; patients with tumor expression of FR-alpha >1639 amol/ug had longer OS than patients with lower FR- alpha levels. TYMS expression <150 amol/ug was similarly predictive of OS. ERCC1 clustered with the DNA damage markers TYMS and pi 6 in poor responders, while patients expressing FR-alpha and E-cadherin had distinctly better survival outcomes.

[0052] It should be apparent to those skilled in the art that many more modifications besides those already described are possible without departing from the inventive concepts herein. The inventive subject matter, therefore, is not to be restricted except in the scope of the appended claims. Moreover, in interpreting both the specification and the claims, all terms should be interpreted in the broadest possible manner consistent with the context. In particular, the terms "comprises" and "comprising" should be interpreted as referring to elements, components, or steps in a non-exclusive manner, indicating that the referenced elements, components, or steps may be present, or utilized, or combined with other elements, components, or steps that are not expressly referenced. Where the specification claims refers to at least one of something selected from the group consisting of A, B, C .... and N, the text should be interpreted as requiring only one element from the group, not A plus N, or B plus N, etc.

Claims (23)

CLAIMS What is claimed is:

1. A method of predicting at least one of overall survival and progression free survival for a patient with non-small cell lung cancer subject to treatment with cisplatin and

pemetrexed, the method comprising:

obtaining quantitative mass spectroscopic data for ERCC1 in a patient sample and classifying the quantitative data as 'not detectable' when the quantitative data is below a threshold value, and as 'detectable' when the quantitative data is above a threshold value; and

updating or generating a patient record as having an improved overall survival and/or progression free survival after treatment with cisplatin/pemetrexed when the quantitative data are 'not detectable'.

2. The method of claim 1 further comprising a step of obtaining additional quantitative mass spectroscopic data for at least one of E-cadherin, HER2, TITFl, MSLN, KRT7, FR-alpha, HER3, FPGS, and ROS1 in the patient sample.

3. The method of claim 2 wherein the additional quantitative mass spectroscopic data are used to predict progression free survival.

4. The method of claim 3 wherein the progression free survival is increased when

expression of the at least one of E-cadherin, HER2, TITFl, MSLN, KRT7, FR-alpha, HER3, FPGS, and ROS1 is increased.

5. The method of any one of the preceding claims wherein the patient sample is selected from the group consisting of a fresh biopsy material, a frozen biopsy sample, and a formalin fixed paraffin embedded sample.

6. The method of any one of the preceding claims wherein the quantitative mass

spectroscopic data is obtained from selected reaction monitoring mass spectroscopy.

7. The method of any one of the preceding claims wherein the threshold value is between two and five times of a standard deviation of background signal.

8. The method of any one of the preceding claims wherein the threshold value is equal or less than lfmol.

9. The method of any one of the preceding claims wherein the patient record is an electronic record.

10. The method of any one of the preceding claims wherein the patient record is updated to include a treatment recommendation to administer cisplatin/pemetrexed.

11. The method of claim 1 wherein the patient sample is selected from the group consisting of a fresh biopsy material, a frozen biopsy sample, and a formalin fixed paraffin embedded sample.

12. The method of claim 1 wherein the quantitative mass spectroscopic data is obtained from selected reaction monitoring mass spectroscopy.

13. The method of claim 1 wherein the threshold value is between two and five times of a standard deviation of background signal.

14. The method of claim 1 wherein the threshold value is equal or less than lfmol.

15. The method of claim 1 wherein the patient record is an electronic record.

16. The method of claim 1 wherein the patient record is updated to include a treatment

recommendation to administer cisplatin/pemetrexed.

17. Use of quantitative mass spectroscopic data for ERCCl in a patient sample to predict at least one of overall survival and progression free survival of a patient with non-small cell lung cancer subject to treatment with cisplatin and pemetrexed,

wherein the quantitative data is classified as 'not detectable' when the quantitative data is below a threshold value, and as 'detectable' when the quantitative data is above a threshold value, and

wherein the patient is determined as having an improved overall survival and/or progression free survival after treatment with cisplatin/pemetrexed when the quantitative data are 'not detectable'.

18. The use of claim 17 further comprising use of expression of at least one of E-cadherin, HER2, TITF1, MSLN, KRT7, FR-alpha, HER3, FPGS, and ROS1 in the patient sample to predict progression free survival of the patient.

19. The use of claim 18 wherein the progression free survival is increased when expression of the at least one of E-cadherin, HER2, TITF1, MSLN, KRT7, FR-alpha, HER3, FPGS, and ROS 1 is increased.

20. The use of any one of claims 17-19 wherein the patient sample is selected from the group consisting of a fresh biopsy material, a frozen biopsy sample, and a formalin fixed paraffin embedded sample.

21. The use of any one of claims 17-20 wherein the quantitative mass spectroscopic data is obtained from selected reaction monitoring mass spectroscopy.

22. The use of any one of claims 17-21 wherein the threshold value is between two and five times of a standard deviation of background signal.

23. The use of any one of claims 17-22 wherein the threshold value is equal or less than

lfmol.