Monitoring response to immunotherapy using cell-free DNA fragmentomes
Monitoring response to immunotherapy using cell-free DNA fragmentomes
1 2 1 1 2 1 2 2 2 1 1 Bahar Alipanahi, Lavanya Sivapalan, Jamie Medina, Zachary L. Skidmore, Paola Ghanem, Erica Peters, Gavin Pereira, Nisha Rao, Kavya Velliangiri, Alissa Konicki, Stephen Cristiano, 1 1 1 2 1 1 3 2 2 1 1 Laurel Millberg, Jacob Carey, Keith Lumbard, Noushin Niknafs, Bryan Chesnick, Jennifer Tom, Alessandro Leal, Benjamin Levy, Patrick Forde, Peter Bach, Nicholas C. Dracopoli, Robert B. 2 2 1 2 Scharpf, Victor Velculescu, Lorenzo Rinaldi, Valsamo Anagnostou
1 2 3 Delfi Diagnostics, Inc., Baltimore, MD, USA; The Sidney Kimmel Comprehensive Cancer Center, Johns Hopkins University School of Medicine; NYU Grossman School of Medicine
BACKGROUND
● The rapid detection of disease progression in patients receiving immune checkpoint inhibitors (ICIs) is challenging given the lack of reliable biomarkers of clinical response. ● Current targeted next-generation sequencing
● Current targeted next-generation sequencing cfDNA assays are costly and require a biomarker or prior knowledge about the mutations the tumor harbors.
● Here, we demonstrate the utility of DELFI Tumor Fraction (DELFI-TF), a tumor- and mutation-independent cfDNA fragmentome approach to monitor treatment response in patients with metastatic non-small cell lung cancer (mNSCLC).
METHODS
● In addition, a cohort consisting of 47 longitudinal samples obtained from 15 mNSCLC patients undergoing treatment with both immunotherapy and chemotherapy was utilized to show the concordance with MAF and RECIST (Multi-Treatment cohort). ● Plasma-derived cfDNA was processed with whole
● Plasma-derived cfDNA was processed with whole genome sequencing (WGS) at low coverage (~4x). ● Circulating tumor burden was quantified as the
● Matched white blood cells were used to filter out germline and clonal hematopoietic variants. ● DELFI-TF, a random forest regression model
● DELFI-TF, a random forest regression model trained on MAF data from longitudinal blood samples of stage IV colorectal cancer patients (van 't Erve et. al. under review), was applied to predict the ctDNA fraction amongst samples. ● The accuracy of DELFI-TF was assessed using
● The accuracy of DELFI-TF was assessed using holdout validation in the ICI and Multi-Treatment cohorts.
● Changes in DELFI-TF and maxMAF at all consecutive timepoints in the ICI cohort were highly correlated (n=215; r=0.9; Pearson); Figure 4. ● DELFI-TF dynamics are consistent with treatment
● In both ICI and Multi-Treatment cohorts, DELFI-TF scores were strongly correlated with maxMAF (n=324, r=0.94, p<0.001, Pearson), (n=47, r=0.94, p<0.001, Pearson) respectively; Figure 2, Figure 3. ● Changes in DELFI-TF and maxMAF at all
RESULTS
TABLE 1. Participant and disease characteristics
| Characteristic | ICI |
|---|---|
| Participants(samples) | 109(324) |
| Median Age | 68(36-91) |
| Sex,n(%) | |
| Male | 59(54%) |
| Female | 50(46%) |
| Stage,n(%) | |
| StageIII | 9(8%) |
| StageIV | 100(92%) |
| Histology,n(%) | |
| Adenocarcinoma | 84(77%) |
| Squamous cell | 23(21%) |
| Large cell carcinoma | 1(1%) |
| Unknown histology | 1(1%) |
| Prior treatment,n(%) | |
| Yes | 72(66%) |
| No | 37(34%) |
| Treatment,n(%) | |
| Immunotherapy only | 80(73%) |
| Immunotherapy+ chemotherapy | 29(27%) |
DELFI-TF scores strongly correlate with maxMAF
● DELFI-TF was applied to all longitudinal samples of the ICI and Multi-Treatment cohorts.
● Figure 2 and Figure 3 illustrate the correlation between DELFI-TF and maxMAF in the ICI cohort (n=324, r = 0.94, p < 0.001, Pearson correlation) and the Multi-Treatment cohort (n=47, r = 0.94, p < 0.001, Pearson correlation). Axis are scaled using a square root transform. ●
DELFI-TF dynamics have strong correlation with maxMAF dynamics and are consistent with the clinical outcomes ● DELFI-TF change between all consecutive timepoints were computed in the ICI cohort. Figure 4
Figure 2. Correlation between DELFI-TF and maxMAF in the ICI cohort.
● DELFI-TF change between all consecutive timepoints were computed in the ICI cohort. Figure 4 illustrates a strong correlation between DELFI-TF and maxMAF changes. Axis are scaled using a square root transform.
root transform. ● DELFI-TF changes are consistent with clinical outcomes. Figure 5 displays DELFI-TF in longitudinal samples of 3 patients experiencing Progressive Disease (PD), Stable Disease (SD) and Partial/Complete Response (PR/CR) during the treatment.
Correlation between DELFI-TF changes and Figure 4. maxMAF changes across all consecutive Figure 5. timepoints in the ICI cohort. (n=215, r = 0.9, p < 0.001, Pearson correlation).
Figure 1A. Patients with liquid biopsy draws at baseline and after treatment with ICI are evaluated for cell-free DNA fragmentation patterns and mutant allele fraction (MAF).
Progression-free survival (PFS) at baseline; Figure 7. DELFI-TF and maxMAF stratified by the median in the ICI cohort. (n=109, p < 0.001 for DELFI-TF and maxMAF, Log-Rank).
Patients with progressive disease have higher DELFI-TF as well as maxMAF Figure 6. values at baseline compared to patients with non-progressive disease in the ICI cohort.
DELFI-TF scores at baseline are prognostic ● Patients who experienced disease progression, more often exhibited increased DELFI-TF values at baseline timepoints than patients who never presented with disease progression; (p = 0.027 and p = 0.003 for DELFI-TF and maxMAF respectively; Wilcoxon); Figure 6. ● Patients with DELFI-TF below the median at the baseline timepoint experienced significantly longer PFS/OS than patients with high DELFI-TF (Figure 7, Figure 8
DELFI-TF scores at baseline are prognostic ● Patients who experienced disease progression, more often exhibited increased DELFI-TF values at baseline timepoints than patients who never presented with