Identification of small molecule aggregators from large compound libraries by support vector machines

Small molecule aggregators non-specifically inhibit multiple unrelated proteins, rendering them therapeutically useless. They frequently appear as false hits and thus need to be eliminated in high-throughput screening campaigns. Computational methods have been explored for identifying aggregators, which have not been tested in screening large compound libraries. We used 1319 aggregators and 128,325 non-aggregators to develop a support vector machines (SVM) aggregator identification model, which was tested by four methods. The first is five fold cross-validation, which showed comparable aggregator and significantly improved non-aggregator identification rates against earlier studies. The second is the independent test of 17 aggregators discovered independently from the training aggregators, 71% of which were correctly identified. The third is retrospective screening of 13M PUBCHEM and 168K MDDR compounds, which predicted 97.9% and 98.7% of the PUBCHEM and MDDR compounds as non-aggregators. The fourth is retrospective screening of 5527 MDDR compounds similar to the known aggregators, 1.14% of which were predicted as aggregators. SVM showed slightly better overall performance against two other machine learning methods based on five fold cross-validation studies of the same settings. Molecular features of aggregation, extracted by a feature selection method, are consistent with published profiles. SVM showed substantial capability in identifying aggregators from large libraries at low false-hit rates.

Heterogeneity of epidermal growth factor receptor (EGFR) expression and variation in immunohistochemistry (IHC) testing may affect access to EGFR-targeted therapy in patients with advanced colorectal cancer (CRC)

10104

Introduction: Cetuximab, an IgG1 MAb directed at the EGFR, has demonstrated benefit in pts with CRC and is currently indicated in EGFR expressing CRC. However, previous reports show heterogeneous EGFR expression between anatomic sites and suggest that results from IHC testing vary between labs. To assess this variability and the resultant impact on pt access to EGFR-targeted therapy, we compared EGFR results reported to clinicians by hospital and reference labs with those obtained at a single central lab. Methods: Formalin-fixed, paraffin-embeded CRC specimens submitted through the Targeted Diagnostics Advocacy Program (tdap) were assayed using the Dako EGFR pharmDx kit and scored using the FDA-approved package insert. If the initial specimen tested EGFR negative (−), additional specimens were requested for assay. Prior EGFR results, lab, and specimen ID were documented through review of anatomic pathology reports. Results: Of 332 evaluable CRC pts tested, 245 pts (74%) were previously reported as EGFR(−). After retesting, 167 (68%) pts were found to be EGFR The originally tested tumor block was retested in 82 pts and 41 pts (50%) were found to be EGFR (+). Retest results were similar for both hospital and reference labs. Of 66 pts with >1 specimen from the same anatomic site tested, 21 (32%) were discordant (at least 1 (+) and one (−) result). Results from different anatomic sites were compared for 65 pts, 34 (52%) of which were discordant (+). Overall, 231 pts (70%) tested EGFR(+). Conclusion: This study shows marked heterogeneity in EGFR expression by IHC both within and across anatomic sites from the same pt as well as variability in both hospital and reference pathology labs. This variability undoubtedly has an adverse impact on pt access to EGFR-targeted therapy. The impact of EGFR heterogeneity may be reduced through the assay of multiple specimens.

[Table: see text]

[Table: see text]

A phase II study to optimize dasatinib dosing in metastatic breast cancer patients using real-time pharmacodynamic tissue and urinary biomarkers

Abstract #3123

Background: In order to optimize drug efficacy and patient selection for targeted agents such as dasatinib, a tyrosine kinase inhibitor, there is a need for adaptive trial methodologies and validated marker endpoints. Predictive and pharmacodynamic (PD) biomarkers of dasatinib activity will be crucial for individualized drug dosing and enrichment of patient populations treated with the drug. Multiple potential predictive and PD markers exist for dasatinib including active/total Src, active/total EphA2, and downstream targets of Src such as focal adhesion kinase (FAK) and paxillin (pax). Given Src's role in bone modeling, markers of bone resorption may also serve as important biomarkers. We designed a phase II trial in metastatic breast cancer (MBC) wherein real-time assessment of these potential biomarkers is used to optimize the dose and anti-tumor effects of dasatinib.
 Material and methods: Key eligibility includes patient with MBC, ECOG 0-1, unlimited prior therapies, and biopsiable tumor. Patients with bone-only MBC were excluded due to tissue quality required for PD analysis. Metastatic biopsies at baseline and week 4 of dasatinib therapy were analyzed using quantitative immunohistochemistry (IHC) (measured in optical densitometry (OD) units) for the following markers: phospho-Src (p-Src), phospho-FAK (p-FAK) and phospho-pax (p-pax). For patients who tolerated the starting dose of dasatinib (50 mg bid), and displayed suboptimal Src inhibition (<80% inhibition of phosphorylation of either biomarker), dasatinib dosing was escalated at week 4 to 70 mg bid. Urinary N-telopeptide (NTX), a marker of bone resorption, was measured monthly.
 Results: Since 12/2007, 12 patients have enrolled and 8 have evaluable, paired metastatic biopsies. Of the other 4 patients, 2 withdrew due to toxicity, 1 voluntarily withdrew, and 1 patient has not yet reached week 4. There has been one case of dyspnea related to possible drug toxicity. Other grade 3/4 toxicities are as follows: anorexia (3), pleural effusion (1), DVT (1). All eligible patients underwent dasatinib dose escalation at week 4 based on their individual tumor biomarker results. All tumors displayed some level of Src inhibition but none of the tumors reached the pre-defined "optimal" level of Src inhibition at week 4; the median changes in tissue biomarker levels are as follows: p-FAK -20%, p-pax -13%, p-Src -9%, and urinary NTX level -11%.
 Conclusions: Our initial analysis of sequential tumor biopsies collected in a phase II trial of single-agent dasatinib in MBC illustrates that real-time biomarkers will both optimize the dosing of targeted agents and define potential on- and off-target drug effects. Accrual is ongoing and updated results of all biomarker endpoints as well as efficacy and toxicity data will be presented.

Citation Information: Cancer Res 2009;69(2 Suppl):Abstract nr 3123.