Srivatsan Raghavan
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Srivatsan Raghavan has a total of 533 co-authors. Here are the top co-authors by number of shared publications:
- Brian M. Wolpin
Dana-Farber Cancer Institute, Boston, MA
181 shared publications - Andrew J. Aguirre
Department of Medical Oncology, Dana-Farber Cancer Institute, Boston, MA 02215, USA
180 shared publications - Jonathan Andrew Nowak
Program in MPE Molecular Pathological Epidemiology, Department of Pathology, Brigham and Women’s Hospital, Boston, Massachusetts, USA
176 shared publications - James M. Cleary
Dana-Farber Cancer Institute 1 Boston, MA
169 shared publications - Kevin S. Kapner
Department of Medical Oncology, Dana-Farber Cancer Institute, Boston, MA 02215, USA
163 shared publications - Kimberly Perez
Dana-Farber Cancer Institute, Boston, MA
156 shared publications
Latest Publications
- Open Access
- 9 November 2026
- Published by Springer Nature in Nature Machine Intelligence
- p. 1-17
Multimodal cell foundation models require synergistic integration beyond alignment
Integrating measurements across modalities and scales into unified quantitative frameworks of cellular function is a central goal in computational biology. Although powerful unimodal foundation models exist, the scarcity of large-scale paired data limits joint multimodal training. This constraint favours compositional foundation models, which fuse frozen unimodal experts through a learned interface. However, it remains unclear when such fusion provides task-relevant information beyond the strongest unimodal representation rather than merely aggregating redundant signal. Here we introduce Synergistic Information Score, a metric grounded in partial information decomposition that quantifies information gain arising only from cross-modal interactions. Extending theory from self-supervised learning, we show that standard alignment-based objectives with frozen encoders collapse to detect linear redundancies, which constrains their ability to capture nonlinear synergistic states. This distinction is especially important for tasks that connect tissue morphology with gene expression. Benchmarking ten fusion methods on spatial transcriptomics datasets, we demonstrate that tasks dominated by linear redundancies are well served by unimodal baselines, whereas complex niche definitions benefit from synergy-aware integration. Finally, we perform a scaling analysis and show that fine-tuning a dominant unimodal expert is the most sample-efficient strategy for standard tasks. Together, these results suggest that progress in multimodal modelling of cellular state requires shifting from alignment-based objectives to synergy-driven integration that exploits complementary cross-modal signal.
- 9 November 2026
- Published by Springer Nature in Nature Machine Intelligence
- p. 1-17
Open AccessMultimodal cell foundation models require synergistic integration beyond alignment
Integrating measurements across modalities and scales into unified quantitative frameworks of cellular function is a central goal in computational biology. Although powerful unimodal foundation models exist, the scarcity of large-scale paired data limits joint multimodal training. This constraint favours compositional foundation models, which fuse frozen unimodal experts through a learned interface. However, it remains unclear when such fusion provides task-relevant information beyond the strongest unimodal representation rather than merely aggregating redundant signal. Here we introduce Synergistic Information Score, a metric grounded in partial information decomposition that quantifies information gain arising only from cross-modal interactions. Extending theory from self-supervised learning, we show that standard alignment-based objectives with frozen encoders collapse to detect linear redundancies, which constrains their ability to capture nonlinear synergistic states. This distinction is especially important for tasks that connect tissue morphology with gene expression. Benchmarking ten fusion methods on spatial transcriptomics datasets, we demonstrate that tasks dominated by linear redundancies are well served by unimodal baselines, whereas complex niche definitions benefit from synergy-aware integration. Finally, we perform a scaling analysis and show that fine-tuning a dominant unimodal expert is the most sample-efficient strategy for standard tasks. Together, these results suggest that progress in multimodal modelling of cellular state requires shifting from alignment-based objectives to synergy-driven integration that exploits complementary cross-modal signal.
- Open Access
- 3 November 2026
- Published by American Association for Cancer Research (AACR) in Cancer Research Communications
A Phase 1 Trial of Low-dose Gemcitabine Combined with ATR Inhibitor Elimusertib for Advanced Solid Tumors
Purpose: Preclinical data suggest that the anti-cancer efficacy of ATR inhibitors can be synergistically enhanced by low-dose gemcitabine. To test the clinical feasibility of this regimen, we conducted a phase 1 trial of low-dose gemcitabine combined with the ATR inhibitor elimusertib (BAY1895344). Patients and Methods: This phase 1 study evaluated the combination of gemcitabine and elimusertib in patients with previously treated, incurable solid tumors. The primary objective was to assess the safety and tolerability of this regimen and to determine the maximum tolerated dose (MTD). Secondary objectives included assessment of objective response rate (ORR) per RECIST v.1.1 and characterization of the pharmacokinetic (PK) profile. Results: Fourteen patients were enrolled; most patients had pancreatic cancer (n=6) or ovarian cancer (n=6). The median age was 63.5 years of age, and the median number of prior therapies was 2. The regimen caused significant toxicity, and four patients (28%) discontinued the trial due to treatment-related adverse events. The trial was stopped prior to establishing an MTD due to the pharmaceutical company's decision to halt development of the agent. A confirmed partial response was observed in one patient with platinum-refractory ovarian cancer harboring a gain in CCNE1. Another patient with pancreatic adenocarcinoma harboring biallelic ATM loss remained on trial for 9.5 months with stable disease, with an 85% decrease in CA19-9. Conclusions: The combination of low-dose gemcitabine combined with elimusertib was not tolerable. Although evidence of anti-tumor activity was observed, significant improvements in tolerability are required for future ATR inhibitor combinations to be clinically feasible.
- 3 November 2026
- Published by American Association for Cancer Research (AACR) in Cancer Research Communications
Open AccessA Phase 1 Trial of Low-dose Gemcitabine Combined with ATR Inhibitor Elimusertib for Advanced Solid Tumors
Purpose: Preclinical data suggest that the anti-cancer efficacy of ATR inhibitors can be synergistically enhanced by low-dose gemcitabine. To test the clinical feasibility of this regimen, we conducted a phase 1 trial of low-dose gemcitabine combined with the ATR inhibitor elimusertib (BAY1895344). Patients and Methods: This phase 1 study evaluated the combination of gemcitabine and elimusertib in patients with previously treated, incurable solid tumors. The primary objective was to assess the safety and tolerability of this regimen and to determine the maximum tolerated dose (MTD). Secondary objectives included assessment of objective response rate (ORR) per RECIST v.1.1 and characterization of the pharmacokinetic (PK) profile. Results: Fourteen patients were enrolled; most patients had pancreatic cancer (n=6) or ovarian cancer (n=6). The median age was 63.5 years of age, and the median number of prior therapies was 2. The regimen caused significant toxicity, and four patients (28%) discontinued the trial due to treatment-related adverse events. The trial was stopped prior to establishing an MTD due to the pharmaceutical company's decision to halt development of the agent. A confirmed partial response was observed in one patient with platinum-refractory ovarian cancer harboring a gain in CCNE1. Another patient with pancreatic adenocarcinoma harboring biallelic ATM loss remained on trial for 9.5 months with stable disease, with an 85% decrease in CA19-9. Conclusions: The combination of low-dose gemcitabine combined with elimusertib was not tolerable. Although evidence of anti-tumor activity was observed, significant improvements in tolerability are required for future ATR inhibitor combinations to be clinically feasible.
- Open Access
Single-cell perturbation prediction near experimental reproducibility with GENESIS
Single-cell perturbation atlases now measure cellular responses across millions of cells, but the number of possible combinations of biological contexts, perturbations, doses and other experimental variables make exhaustive measurements infeasible. Here, we present GENESIS (Generative ENErgy-based Single-cell Intervention Simulator), a generative framework that predicts single-cell response distributions directly from the covariates defining an experimental condition. GENESIS uses a lightweight neural network trained with the energy score, a strictly proper scoring rule for distributions. We introduce an experiment-calibrated evaluation framework in which prediction error is interpreted relative to the reproducibility of repeated measurements of the same condition. Across the Tahoe-100M small-molecule atlas and a large PBMC cytokine atlas, GENESIS significantly outperforms existing generative models and statistical baselines and approaches experimental reproducibility on independently repeated conditions. Importantly, GENESIS remains accurate when trained on only a fraction of the available experimental conditions or cells, and generated profiles recover experimentally observed context-dependent transcriptional programs. At comparable training-cell budgets, GENESIS achieves more accurate predictions when measurements span more experimental conditions rather than profiling fewer conditions more deeply. Together, these findings suggest a complementary strategy for perturbation studies: broader, lower-depth measurements can expand experimental coverage, while GENESIS augments sparsely sampled populations and predicts responses in unmeasured conditions.
Open AccessSingle-cell perturbation prediction near experimental reproducibility with GENESIS
Single-cell perturbation atlases now measure cellular responses across millions of cells, but the number of possible combinations of biological contexts, perturbations, doses and other experimental variables make exhaustive measurements infeasible. Here, we present GENESIS (Generative ENErgy-based Single-cell Intervention Simulator), a generative framework that predicts single-cell response distributions directly from the covariates defining an experimental condition. GENESIS uses a lightweight neural network trained with the energy score, a strictly proper scoring rule for distributions. We introduce an experiment-calibrated evaluation framework in which prediction error is interpreted relative to the reproducibility of repeated measurements of the same condition. Across the Tahoe-100M small-molecule atlas and a large PBMC cytokine atlas, GENESIS significantly outperforms existing generative models and statistical baselines and approaches experimental reproducibility on independently repeated conditions. Importantly, GENESIS remains accurate when trained on only a fraction of the available experimental conditions or cells, and generated profiles recover experimentally observed context-dependent transcriptional programs. At comparable training-cell budgets, GENESIS achieves more accurate predictions when measurements span more experimental conditions rather than profiling fewer conditions more deeply. Together, these findings suggest a complementary strategy for perturbation studies: broader, lower-depth measurements can expand experimental coverage, while GENESIS augments sparsely sampled populations and predicts responses in unmeasured conditions.
- Top Cited PapersOpen Access
- 5 August 2026
- Published by Springer Nature in Nature
- Vol. 657 (8132) , 789-799
A dependency map enhanced with next-generation 3D cancer models
- 5 August 2026
- Published by Springer Nature in Nature
- Vol. 657 (8132) , 789-799
Top Cited PapersOpen AccessA dependency map enhanced with next-generation 3D cancer models
- Open Access
- 17 June 2026
- Published by American Association for Cancer Research (AACR)
Data from A Phase II Trial of Niraparib in Patients with Advanced Pancreatic Cancer Harboring Pathogenic Variants in ATM, BRCA1, BRCA2, PALB2, and CHEK2
Purpose:PARP inhibition has demonstrated efficacy in patients with platinum-sensitive pancreatic cancer with germline BRCA1/2 pathogenic variants (PV). Whether PARP inhibitors might be effective in a broader population of patients with pancreatic cancer remains under investigation.Patients and Methods:This multicenter, open-label phase II trial (NCT03601923) enrolled patients with advanced pancreatic cancers who harbored germline or somatic BRCA1, BRCA2, PALB2, ATM, and/or CHEK2 PVs. Patients with prior progression on platinum-based therapy were excluded. Patients were treated with niraparib 200 or 300 mg once daily, with their initial dose determined by weight and platelet count. The primary endpoint was 6-month progression-free survival (PFS).Results:Thirty-two patients [10 women (31%); median age: 67 years] were enrolled. Patients had a PV in at least one of the following genes: ATM (n = 14), BRCA2 (n = 10), PALB2 (n = 3), CHEK2 (n = 4), or BRCA1 (n = 2). The 6-month PFS was 25% [90% confidence interval (CI), 13%–41%] for the overall population, exceeding the preestablished threshold of 17%. The median PFS for the entire population was 2 months (95% CI, 1.4–3.8), and the objective response rate was 14% (95% CI, 4%–33%). All six patients with ≥6 months of PFS and evaluable tumor zygosity had biallelic inactivation of a DNA repair gene. Three patients with biallelic inactivation of ATM and no progression on prior chemotherapy received niraparib for more than 1 year.Conclusions:Niraparib demonstrated clinically meaningful benefit in a subset of patients. The prolonged PFS observed in patients with biallelic ATM loss warrants further investigation.
- 17 June 2026
- Published by American Association for Cancer Research (AACR)
Open AccessData from A Phase II Trial of Niraparib in Patients with Advanced Pancreatic Cancer Harboring Pathogenic Variants in ATM, BRCA1, BRCA2, PALB2, and CHEK2
Purpose:PARP inhibition has demonstrated efficacy in patients with platinum-sensitive pancreatic cancer with germline BRCA1/2 pathogenic variants (PV). Whether PARP inhibitors might be effective in a broader population of patients with pancreatic cancer remains under investigation.Patients and Methods:This multicenter, open-label phase II trial (NCT03601923) enrolled patients with advanced pancreatic cancers who harbored germline or somatic BRCA1, BRCA2, PALB2, ATM, and/or CHEK2 PVs. Patients with prior progression on platinum-based therapy were excluded. Patients were treated with niraparib 200 or 300 mg once daily, with their initial dose determined by weight and platelet count. The primary endpoint was 6-month progression-free survival (PFS).Results:Thirty-two patients [10 women (31%); median age: 67 years] were enrolled. Patients had a PV in at least one of the following genes: ATM (n = 14), BRCA2 (n = 10), PALB2 (n = 3), CHEK2 (n = 4), or BRCA1 (n = 2). The 6-month PFS was 25% [90% confidence interval (CI), 13%–41%] for the overall population, exceeding the preestablished threshold of 17%. The median PFS for the entire population was 2 months (95% CI, 1.4–3.8), and the objective response rate was 14% (95% CI, 4%–33%). All six patients with ≥6 months of PFS and evaluable tumor zygosity had biallelic inactivation of a DNA repair gene. Three patients with biallelic inactivation of ATM and no progression on prior chemotherapy received niraparib for more than 1 year.Conclusions:Niraparib demonstrated clinically meaningful benefit in a subset of patients. The prolonged PFS observed in patients with biallelic ATM loss warrants further investigation.
- Open Access
- 17 June 2026
- Published by American Association for Cancer Research (AACR)
Supplementary Figure S1 from A Phase II Trial of Niraparib in Patients with Advanced Pancreatic Cancer Harboring Pathogenic Variants in ATM, BRCA1, BRCA2, PALB2, and CHEK2
Supplementary Figure S1: CONSORT diagram indicating patient disposition.
- 17 June 2026
- Published by American Association for Cancer Research (AACR)
Open AccessSupplementary Figure S1 from A Phase II Trial of Niraparib in Patients with Advanced Pancreatic Cancer Harboring Pathogenic Variants in ATM, BRCA1, BRCA2, PALB2, and CHEK2
Supplementary Figure S1: CONSORT diagram indicating patient disposition.
- Open Access
- 17 June 2026
- Published by American Association for Cancer Research (AACR)
Supplemental Table 6 from Mechanisms of Resistance to Oncogenic KRAS Inhibition in Pancreatic Cancer
Mutation and sequence alterations from whole exome sequencing of in vitro models of acquired resistance to MRTX1133.
- 17 June 2026
- Published by American Association for Cancer Research (AACR)
Open AccessSupplemental Table 6 from Mechanisms of Resistance to Oncogenic KRAS Inhibition in Pancreatic Cancer
Mutation and sequence alterations from whole exome sequencing of in vitro models of acquired resistance to MRTX1133.
- Open Access
- 17 June 2026
- Published by American Association for Cancer Research (AACR)
Supplemental Table 4 from Mechanisms of Resistance to Oncogenic KRAS Inhibition in Pancreatic Cancer
Supplemental Table 4A: Differential gene expression analysis between MRTX1133 sensitive and resistant KRASG12D cell lines. Supplemental Table 4B: Analysis of Copy number and RPPA datasets between MRTX1133 sensitive and resistant KRASG12D cell lines.
- 17 June 2026
- Published by American Association for Cancer Research (AACR)
Open AccessSupplemental Table 4 from Mechanisms of Resistance to Oncogenic KRAS Inhibition in Pancreatic Cancer
Supplemental Table 4A: Differential gene expression analysis between MRTX1133 sensitive and resistant KRASG12D cell lines. Supplemental Table 4B: Analysis of Copy number and RPPA datasets between MRTX1133 sensitive and resistant KRASG12D cell lines.
- Open Access
- 17 June 2026
- Published by American Association for Cancer Research (AACR)
Supplementary Figure S6 from A Phase II Trial of Niraparib in Patients with Advanced Pancreatic Cancer Harboring Pathogenic Variants in ATM, BRCA1, BRCA2, PALB2, and CHEK2
Supplementary Figure S6: Genomic characterization of BRCA2 reversion mutations
- 17 June 2026
- Published by American Association for Cancer Research (AACR)
Open AccessSupplementary Figure S6 from A Phase II Trial of Niraparib in Patients with Advanced Pancreatic Cancer Harboring Pathogenic Variants in ATM, BRCA1, BRCA2, PALB2, and CHEK2
Supplementary Figure S6: Genomic characterization of BRCA2 reversion mutations
- Open Access
- 17 June 2026
- Published by American Association for Cancer Research (AACR)
Supplementary Figure S2 from A Phase II Trial of Niraparib in Patients with Advanced Pancreatic Cancer Harboring Pathogenic Variants in ATM, BRCA1, BRCA2, PALB2, and CHEK2
Supplementary Figure S2: Kaplan-Meier estimates of overall survival (OS) in patients treated with niraparib across the entire trial population
- 17 June 2026
- Published by American Association for Cancer Research (AACR)
Open AccessSupplementary Figure S2 from A Phase II Trial of Niraparib in Patients with Advanced Pancreatic Cancer Harboring Pathogenic Variants in ATM, BRCA1, BRCA2, PALB2, and CHEK2
Supplementary Figure S2: Kaplan-Meier estimates of overall survival (OS) in patients treated with niraparib across the entire trial population