Validation of Immunotherapy Response Score as Predictive of Pan-solid Tumor Anti-PD-1/PD-L1 Benefit.

Immunotherapy response score (IRS) integrates tumor mutation burden (TMB) and quantitative expression biomarkers to predict anti-PD-1/PD-L1 [PD-(L)1] monotherapy benefit. Here, we evaluated IRS in additional cohorts. Patients from an observational trial (NCT03061305) treated with anti-PD-(L)1 monotherapy were included and assigned to IRS-High (-H) versus -Low (-L) groups. Associations with real-world progression-free survival (rwPFS) and overall survival (OS) were determined by Cox proportional hazards (CPH) modeling. Those with available PD-L1 IHC treated with anti-PD-(L)1 with or without chemotherapy were separately assessed. Patients treated with PD-(L)1 and/or chemotherapy (five relevant tumor types) were assigned to three IRS groups [IRS-L divided into IRS-Ultra-Low (-UL) and Intermediate-Low (-IL), and similarly assessed]. In the 352 patient anti-PD-(L)1 monotherapy validation cohort (31 tumor types), IRS-H versus IRS-L patients had significantly longer rwPFS and OS. IRS significantly improved CPH associations with rwPFS and OS beyond microsatellite instability (MSI)/TMB alone. In a 189 patient (10 tumor types) PD-L1 IHC comparison cohort, IRS, but not PD-L1 IHC nor TMB, was significantly associated with anti-PD-L1 rwPFS. In a 1,103-patient cohort (from five relevant tumor types), rwPFS did not significantly differ in IRS-UL patients treated with chemotherapy versus chemotherapy plus anti-PD-(L)1, nor in IRS-H patients treated with anti-PD-(L)1 versus anti-PD-(L)1 + chemotherapy. IRS associations were consistent across subgroups, including both Europeans and non-Europeans. These results confirm the utility of IRS utility for predicting pan-solid tumor PD-(L)1 monotherapy benefit beyond available biomarkers and demonstrate utility for informing on anti-PD-(L)1 and/or chemotherapy treatment. Significance: This study confirms the utility of the integrative IRS biomarker for predicting anti-PD-L1/PD-1 benefit. IRS significantly improved upon currently available biomarkers, including PD-L1 IHC, TMB, and MSI status. Additional utility for informing on chemotherapy, anti-PD-L1/PD-1, and anti-PD-L1/PD-1 plus chemotherapy treatments decisions is shown.

Development and validation of an integrative pan-solid tumor predictor of PD-1/PD-L1 blockade benefit.

BACKGROUND: Anti-PD-1 and PD-L1 (collectively PD-[L]1) therapies are approved for many advanced solid tumors. Biomarkers beyond PD-L1 immunohistochemistry, microsatellite instability, and tumor mutation burden (TMB) may improve benefit prediction. METHODS: Using treatment data and genomic and transcriptomic tumor tissue profiling from an observational trial (NCT03061305), we developed Immunotherapy Response Score (IRS), a pan-tumor predictive model of PD-(L)1 benefit. IRS real-world progression free survival (rwPFS) and overall survival (OS) prediction was validated in an independent cohort of trial patients. RESULTS: Here, by Cox modeling, we develop IRS-which combines TMB with CD274, PDCD1, ADAM12 and TOP2A quantitative expression-to predict pembrolizumab rwPFS (648 patients; 26 tumor types; IRS-High or -Low groups). In the 248 patient validation cohort (248 patients; 24 tumor types; non-pembrolizumab PD-[L]1 monotherapy treatment), median rwPFS and OS are significantly longer in IRS-High vs. IRS-Low patients (rwPFS adjusted hazard ratio [aHR] 0.52, p = 0.003; OS aHR 0.49, p = 0.005); TMB alone does not significantly predict PD-(L)1 rwPFS nor OS. In 146 patients treated with systemic therapy prior to pembrolizumab monotherapy, pembrolizumab rwPFS is only significantly longer than immediately preceding therapy rwPFS in IRS-High patients (interaction test p = 0.001). In propensity matched lung cancer patients treated with first-line pembrolizumab monotherapy or pembrolizumab+chemotherapy, monotherapy rwPFS is significantly shorter in IRS-Low patients, but is not significantly different in IRS-High patients. Across 24,463 molecularly-evaluable trial patients, 7.6% of patients outside of monotherapy PD-(L)1 approved tumor types are IRS-High/TMB-Low. CONCLUSIONS: The validated, predictive, pan-tumor IRS model can expand PD-(L)1 monotherapy benefit outside currently approved indications.

Therapies activating the immune system (checkpoint inhibitors) have revolutionized the treatment of patients with advanced cancer, however new molecular tests may better identify patients who could benefit. Using treatment data and clinical molecular test results, we report the development and validation of Immunotherapy Response Score (IRS) to predict checkpoint inhibitor benefit. Across patients with more than 20 advanced cancer types, IRS better predicted checkpoint inhibitor benefit than currently available tests. Data from >20,000 patients showed that IRS identifies ~8% of patients with advanced cancer who may dramatically benefit from checkpoint inhibitors but would not receive them today based on currently available tests. Our approach may help clinicians to decide which patients should receive checkpoint inhibitors to treat their disease.

Analytical and Potential Clinical Performance of Oncomine Myeloid Research Assay for Myeloid Neoplasms.

INTRODUCTION: Next-generation sequencing (NGS) panels have recently been introduced to efficiently detect genetic variations in hematologic malignancies. OBJECTIVES: Our aim was to evaluate the performance of the commercialized Oncomine™ myeloid research assay (OMA) for myeloid neoplasms. METHODS: Certified reference materials and clinical research samples were used, including 60 genomic DNA and 56 RNA samples. NGS was performed using OMA, which enables the interrogation of 40 target genes, 29 gene fusions, and five expression target genes with five expression control genes by the Ion S5 XL Sequencer. The analyzed data were compared with clinical data using karyotyping, reverse transcription polymerase chain reaction (PCR), fluorescence in situ hybridization, Sanger sequencing, customized NGS panel, and fragment analysis. RESULTS: All targets of reference materials were detected except three (two ASXL1 and one CEBPA) mutations, which we had not expected OMA to detect. In clinical search samples, OMA satisfactorily identified DNA variants, including 90 single nucleotide variants (SNVs), 48 small insertions and deletions (indels), and eight FLT3 internal tandem duplications (ITDs) (Kappa agreement 0.938). The variant allele frequencies of SNVs and indels measured by OMA correlated well with clinical data, whereas those of FLT3-ITDs were significantly lower than with fragment analysis (P = 0.008). Together, OMA showed strong ability to identify RNA gene fusions (Kappa agreement 0.961), except one RUNX1-MECOM. The MECOM gene was highly expressed in all five samples with MECOM-associated rearrangements, including inv(3), t(3;3), and t(3;21). CONCLUSION: OMA revealed excellent analytical and potential clinical performance and could be a good replacement for conventional molecular tests.

Exploring the composition of protein-ligand binding sites on a large scale.

The residue composition of a ligand binding site determines the interactions available for diffusion-mediated ligand binding, and understanding general composition of these sites is of great importance if we are to gain insight into the functional diversity of the proteome. Many structure-based drug design methods utilize such heuristic information for improving prediction or characterization of ligand-binding sites in proteins of unknown function. The Binding MOAD database if one of the largest curated sets of protein-ligand complexes, and provides a source of diverse, high-quality data for establishing general trends of residue composition from currently available protein structures. We present an analysis of 3,295 non-redundant proteins with 9,114 non-redundant binding sites to identify residues over-represented in binding regions versus the rest of the protein surface. The Binding MOAD database delineates biologically-relevant "valid" ligands from "invalid" small-molecule ligands bound to the protein. Invalids are present in the crystallization medium and serve no known biological function. Contacts are found to differ between these classes of ligands, indicating that residue composition of biologically relevant binding sites is distinct not only from the rest of the protein surface, but also from surface regions capable of opportunistic binding of non-functional small molecules. To confirm these trends, we perform a rigorous analysis of the variation of residue propensity with respect to the size of the dataset and the content bias inherent in structure sets obtained from a large protein structure database. The optimal size of the dataset for establishing general trends of residue propensities, as well as strategies for assessing the significance of such trends, are suggested for future studies of binding-site composition.

Overcoming sequence misalignments with weighted structural superposition.

An appropriate structural superposition identifies similarities and differences between homologous proteins that are not evident from sequence alignments alone. We have coupled our Gaussian-weighted RMSD (wRMSD) tool with a sequence aligner and seed extension (SE) algorithm to create a robust technique for overlaying structures and aligning sequences of homologous proteins (HwRMSD). HwRMSD overcomes errors in the initial sequence alignment that would normally propagate into a standard RMSD overlay. SE can generate a corrected sequence alignment from the improved structural superposition obtained by wRMSD. HwRMSD's robust performance and its superiority over standard RMSD are demonstrated over a range of homologous proteins. Its better overlay results in corrected sequence alignments with good agreement to HOMSTRAD. Finally, HwRMSD is compared to established structural alignment methods: FATCAT, secondary-structure matching, combinatorial extension, and Dalilite. Most methods are comparable at placing residue pairs within 2 Å, but HwRMSD places many more residue pairs within 1 Å, providing a clear advantage. Such high accuracy is essential in drug design, where small distances can have a large impact on computational predictions. This level of accuracy is also needed to correct sequence alignments in an automated fashion, especially for omics-scale analysis. HwRMSD can align homologs with low-sequence identity and large conformational differences, cases where both sequence-based and structural-based methods may fail. The HwRMSD pipeline overcomes the dependency of structural overlays on initial sequence pairing and removes the need to determine the best sequence-alignment method, substitution matrix, and gap parameters for each unique pair of homologs.

CSAR benchmark exercise of 2010: selection of the protein-ligand complexes.

A major goal in drug design is the improvement of computational methods for docking and scoring. The Community Structure Activity Resource (CSAR) aims to collect available data from industry and academia which may be used for this purpose ( www.csardock.org ). Also, CSAR is charged with organizing community-wide exercises based on the collected data. The first of these exercises was aimed to gauge the overall state of docking and scoring, using a large and diverse data set of protein-ligand complexes. Participants were asked to calculate the affinity of the complexes as provided and then recalculate with changes which may improve their specific method. This first data set was selected from existing PDB entries which had binding data (K(d) or K(i)) in Binding MOAD, augmented with entries from PDB bind. The final data set contains 343 diverse protein-ligand complexes and spans 14 pK(d). Sixteen proteins have three or more complexes in the data set, from which a user could start an inspection of congeneric series. Inherent experimental error limits the possible correlation between scores and measured affinity; Pearson R is limited to ~ 0.91 (Pearson R2 0.83) when fitting to the data set without over parameterizing. Pearson R is limited to ~ 0.83(Pearson R2 ~ 0.70) when scoring the data set with a method trained on outside data [corrected]. The details of how the data set was initially selected, and the process by which it matured to better fit the needs of the community are presented. Many groups generously participated in improving the data set, and this underscores the value of a supportive, collaborative effort in moving our field forward.

Creating an ontology-based human phenotyping system: The Rockefeller University bleeding history experience.

The lack of standardized methods for human phenotyping is a major obstacle in translational science. We have developed a bleeding history phenotyping system comprising an ontology, a questionnaire, a Web-based phenotype recording instrument (PRI), and a database. The ontology facilitates transparency, collaboration, aggregation of data, and data analysis. The integrated system allows investigators worldwide to use the PRI, add their de-identified data to the database, and query the aggregated data. Thus, this system can increase the power to detect genotype-phenotype-environment relationships and help new investigators begin their studies. We anticipate that this approach may be applicable to other disorders.

Differences between high- and low-affinity complexes of enzymes and nonenzymes.

Physical differences in small molecule binding between enzymes and nonenzymes were found through mining the protein-ligand database, Binding MOAD (Mother of All Databases). The data suggest that divergent approaches may be more productive for improving the affinity of ligands for the two classes of proteins. High-affinity ligands of enzymes are much larger than those with low affinity, indicating that the addition of complementary functional groups is likely to improve the affinity of an enzyme inhibitor. However, this process may not be as fruitful for ligands of nonenzymes. High- and low-affinity ligands of nonenzymes are nearly the same size, so modest modifications and isosteric replacement might be most productive. The inherent differences between enzymes and nonenzymes have significant ramifications for scoring functions and structure-based drug design. In particular, nonenzymes were found to have greater ligand efficiencies than enzymes. Ligand efficiencies are often used to indicate druggability of a target, and this finding supports the feasibility of nonenzymes as drug targets. The differences in ligand efficiencies do not appear to come from the ligands; instead, the pockets yield different amino acid compositions despite very similar distributions of amino acids in the overall protein sequences.

Binding MOAD, a high-quality protein-ligand database.

Binding MOAD (Mother of All Databases) is a database of 9836 protein-ligand crystal structures. All biologically relevant ligands are annotated, and experimental binding-affinity data is reported when available. Binding MOAD has almost doubled in size since it was originally introduced in 2004, demonstrating steady growth with each annual update. Several technologies, such as natural language processing, help drive this constant expansion. Along with increasing data, Binding MOAD has improved usability. The website now showcases a faster, more featured viewer to examine the protein-ligand structures. Ligands have additional chemical data, allowing for cheminformatics mining. Lastly, logins are no longer necessary, and Binding MOAD is freely available to all at http://www.BindingMOAD.org.