Browse Articles

Artificial intelligence and machine learning in relapsed/refractory multiple myeloma: A systematic review.

e19500 Background: Relapsed/refractory multiple myeloma (RRMM) remains incurable with heterogeneous outcomes. Traditional prognostic systems (ISS/R-ISS) offer limited discriminatory power and no treatment guidance in heterogeneous RRMM. The heterogeneity of RRMM makes it suitable for AI/ML applications; however, these approaches have not been systematically reviewed. Methods: We systematically searched PubMed, Embase, and Web of Science (till December 2025) for studies applying AI/ML to RRMM populations. Included studies required: (1) AI/ML algorithm application, (2) RRMM patients included or validated, and (3) clinical outcome prediction. Studies were categorized by application domain. Performance metrics (C-index, AUC), validation methods, and clinical applicability were extracted. Results: We identified 20 studies applying AI/ML to >20,000 patients across six RRMM-relevant domains. Survival prediction (n=7): Transformer, Random Forest, and ensemble models achieved C-indices of 0.62–0.82; the EMN-HARMONY model (N=14,345) showed 6-variable score outperforming R-ISS, while POD24 was strongest mortality predictor. CAR T-cell therapy (n=3): The MyCARe logistic regression model stratified 5-month relapse risk (7% vs 27% vs 53%; p<0.001) using 4 clinical variables; single-cell multiomics identified CD4+ cytotoxic T cells as predictors of both response and toxicity. Treatment response (n=4): IsomiR-based classifiers achieved high accuracy (AUC 0.98) for daratumumab response; multiomics network models (mmSYGNAL) outperformed ISS and cytogenetics across disease trajectory. Toxicity prediction (n=3): ML nomograms predicted carfilzomib-associated heart failure (C-index 0.78) and daratumumab-related pneumonia (AUC >0.80) using ECOG, hemoglobin, and comorbidities. Minimal residual disease (MRD) (n=1): AI-based pattern classification of longitudinal MRD significantly stratified PFS (p<10⁻⁷). Imaging radiomics (n=2): 3D-CNN on whole-body MRI (AUC 0.804) and DeepSurv on PET/CT (C-index 0.657) identified prognostic imaging features. External validation: 11/20 studies (55%). Conclusions: AI/ML demonstrates substantial potential for RRMM risk stratification, CAR T optimization, treatment response prediction, and toxicity prevention. Three models (MyCARe, EMN-HARMONY, carfilzomib cardiotoxicity nomogram) use routine clinical variables and are immediately implementable. Standardized external validation and prospective trials are needed for widespread clinical application. Study Method Finding Metric Mosquera 2025 Ensemble ML 6-variables vs R-ISS C-index:0.66 Gagelmann 2024 Logistic MyCARe CAR T relapse score 7% vs 53% relapse Gómez-Martín 2025 IsomiR Daratumumab response AUC:0.98 Qiao 2025 XGBoost carfilzomib cardiotoxicity C:0.78 Martinez-Lopez 2024 AI class MRD dynamics p10⁻⁷ Morita 2024 3D-CNN whole-body-MRI prognosis AUC:0.80

Association of dietary fiber and fructose with progression-free survival (PFS) in metastatic/advanced urothelial cancer (mUC) on immune checkpoint blockade (ICB) independent of correlates of metabolic health.

e16569 Background: Diet can modulate the gut microbiome and may affect ICB efficacy. In a cohort of pts with mUC on ICB, we previously reported associations of longer PFS with high dietary fiber and low dietary fructose. These associations were independent of clinical factors. However, observational studies can be affected by confounding, so we sought to determine if confounding by predictors of metabolic health explained our prior findings. Methods: In a retrospective analysis of a prospectively collected cohort, we leveraged dietary data collected with the Harvard Willett Food Frequency Questionnaire from pts with mUC initiating ICB at Memorial Sloan Kettering to assess for associations between PFS and baseline BMI, daily caloric intake (cal/d), and dietary glucose. These were treated as continuous variables and log transformed if skewed. Associations with PFS were assessed by univariate (UV) & multivariable (MV) Cox proportional hazards regression. Results: From 2/2021-6/2022, 38 pts eligible for analysis enrolled. Median follow-up was 10.4 months with 27 PFS events. Median (with interquartile range) for each variable of interest: fiber 17 g/day (13-22); fructose 15 g/day (12-31); BMI 26.5 kg/m2 (23.4-30.4); glucose 15 g/day (11-25); cal/d 1,597 (1,091-1,939). Visceral metastases were present in 21 (55%) pts, and 5 (13%) had prior ICB. BMI, glucose, and cal/d were not associated with PFS in UV models (BMI HR 0.27, 95% CI 0.02-3.29, p=0.31; glucose HR 1.46, 95% CI 0.74-2.86, p=0.27; cal/d HR 0.74, 95% CI 0.30-1.79, p=0.50), models adjusted for tumor mutational burden, Bellmunt risk factors, and prior ICB (BMI HR 0.34, 95% CI 0.02-4.96; p=0.43; glucose HR 1.3, 95% CI 0.63-2.64, p=0.49; cal/d HR 0.51, 95% CI 0.19-1.40, p=0.19), nor models adjusted for fiber and fructose intake (BMI HR 0.27, 95% CI 0.02-3.11; p=0.29; glucose HR 0.51, 95% CI 0.06-4.61, p=0.55; cal/d HR 0.67, 95% CI 0.13-3.45, p=0.63). Associations of fiber and fructose intake with PFS persisted after adjusting for BMI, glucose, and cal/d (Table 1). Conclusions: In mUC, BMI, cal/d, and dietary glucose were not associated with PFS on ICB. Confounding by these variables did not account for associations of high dietary fiber and low dietary fructose with longer PFS. Hazard ratios (HR) with 95% confidence intervals (CI) for PFS. All MV models included fiber & fructose. Dietary variable UV models Model for fiber & fructose together Adjusted for BMI Adjusted for glucose intake Adjusted for cal/d Fiber, g/day 0.98 (0.94-1.02); p = 0.28 0.90 (0.84-0.97); p = 0.005 0.91 (0.85-0.97); p = 0.004 0.91 (0.84-0.97); p = 0.008 0.91 (0.84-0.996); p = 0.04 Fructose, log(g/day) 1.66 (0.89-3.09); p = 0.11 5.02 (2.06-12.23); p = 0.0004 5.06 (2.08-12.32); p = 0.0004 8.65 (1.16-64.30); p = 0.035 5.44 (2.08-14.18); p = 0.0005

Low-dose intestinal irradiation to enhance the efficacy and prognosis of PD-1 blockade in metastatic non-small cell lung cancer.

2627 Background: Intestinal low-dose irradiation (ILDR) has been shown to enhance the efficacy of immunotherapy in advanced solid tumors by modulating the gut microbiota and metabolism. However, its role in metastatic non-small cell lung cancer (mNSCLC), particularly in the first-line treatment setting, remains unclear. Therefore, this study investigated the impact of intestinal radiation dose on the efficacy and prognosis of programmed cell death protein 1 (PD-1) blockade in patients with mNSCLC. Methods: This multicenter retrospective and prospective study included patients with mNSCLC who received first- or second-line PD-1 inhibitors combined with abdominopelvic radiotherapy between 2018 and 2025. Patients were stratified into three groups according to the mean intestinal radiation dose: <1 Gy, 1–3 Gy, and >3 Gy, and treatment outcomes were compared among groups. In addition, blood and fecal samples prospectively collected were subjected to multi-omics analyses. Results: A total of 301 patients were included in the retrospective analysis, among whom 105 patients (34.9%) had a small intestinal mean radiation dose (SIMRD) <1 Gy, and 84 patients (27.9%) had a SIMRD of 1–3 Gy. Overall, 183 patients (60.8%) received first-line PD-1 blockade, and 118 patients (39.2%) received second-line treatment. The median follow-up time was 27.2 months. The results showed that patients with a SIMRD of 1–3 Gy achieved the highest objective response rate (21.0% vs. 48.8% vs. 6.3%). Compared with the <1 Gy and >3 Gy groups, the 1–3 Gy group also demonstrated significantly prolonged progression-free survival (PFS, 10.2 months) and overall survival (OS, 23.7 months) (P < 0.01), with consistent findings across all subgroup analyses. Compared with the 1–3 Gy group, SIMRD >3 Gy (HR = 4.96, P < 0.001) and SIMRD <1 Gy (HR = 1.90, P < 0.001) were both independent predictors of worse OS. Thirty patients who recevied first-line PD-1 inhibitors combined with abdominopelvic radiotherapy were included in the prospective analyses. With a median follow-up of 17.5 months, patients with a SIMRD of 1–3 Gy achieved the best disease control rate (1-3 Gy vs. <1 Gy vs. >3 Gy: 90.0% vs. 63.7% vs. 33.3%; P = 0.041) and the longest median PFS (1-3 Gy vs. <1 Gy vs. >3 Gy: not reached vs. 9.3 months vs. 5.8 months; P = 0.120). Multi-omics analyses revealed that responders were enriched in Bacillota , Clostridia , and indole-derived metabolites, particularly indole-3-carboxylic acid. Moreover, patients in the 1–3 Gy group exhibited increased circulating macrophage inflammatory protein-3α levels and a reduced proportion of α4β7 + regulatory T cells. Conclusions: ILDR influences the efficacy of PD-1 inhibitor therapy in patients with mNSCLC, with the most pronounced benefit observed when SIMRD is maintained within the 1–3 Gy range. This effect may be mediated through modulation of the gut microbiota–metabolite–immune axis.

Updated analysis of overall survival (OS) with imetelstat (IME) in relapsed/refractory (R/R) myelofibrosis (MF) versus real-world (RW) data, and assessment of RW treatment (tx) patterns.

6573 Background: IME showed significantly longer median OS (mOS) in patients (pts) with Janus kinase inhibitor (JAKi) R/R intermediate-2–risk or high-risk MF in the IMbark trial (NCT02426086) vs closely matched RW pts treated with best available therapy (BAT) after ruxolitinib (RUX; HR, 0.35; P =.0019). The MF tx landscape changed considerably in recent years, potentially impacting OS. We present an updated OS analysis of IMbark pts vs a larger RW pt cohort after extended follow-up. Methods: The updated RW dataset included 126 pts who discontinued RUX and were subsequently treated with BAT at Moffitt Cancer Center between 2010 and 2025. To assess changes in RW tx patterns, baseline/disease characteristics and OS of pts diagnosed before/after 2016 and 2019 were compared. To update the IMbark vs RW OS analysis, a closely matched cohort was identified using IMbark eligibility criteria, including pts diagnosed before 2016 who received RUX but were R/R. The mOS was measured from time of JAKi discontinuation to death or censored at last follow-up. Propensity score approaches using average tx effect for overlap population (ATO) or stabilized inverse probability tx weighting (sIPTW) were implemented to adjust for baseline covariates/prognostic factors that may impact outcomes. Results: The RW tx pattern assessment included 96 pts. Time from diagnosis to start of RUX and duration of RUX tx were significantly shorter after 2016 vs before (Table). Similar changes were observed after 2019 vs before. The mOS in pts diagnosed after 2016 and 2019 was significantly longer vs before. The updated mOS analysis included 59 IMbark pts and 54 closely matched RW pts. With a median follow-up of 46.2 mo for IMbark pts and 48.2 mo for RW pts, mOS (95% CI) was 30.7 mo (25.5-36.9) with IME in IMbark vs 15.4 mo (13.1-30.5) with BAT in RW (HR, 0.512; P= .003) per the unweighted analysis. Propensity-weighted analyses (ATO, sIPTW) had similar results. Conclusions: This updated post hoc analysis confirms a significantly more favorable OS benefit with IME vs BAT in pts with R/R MF and poor prognosis consistent with the previous report (Kuykendall 2021). A significant improvement in OS in RW pts over the last decade was also demonstrated, potentially due to shorter time from diagnosis to start of RUX and earlier switch to next-line JAKi or clinical trial. Evolving tx patterns are key considerations for interpreting future clinical trial outcomes. Clinical trial information: NCT02426086 . Updated RW data by time of diagnosis. Before 2016(N=54) After 2016(N=42) Before 2019(N=69) After 2019(N=27) Time from diagnosis to RUX start (mean), mo 46.7 10.5 39.7 8.2 RUX duration (mean), mo 30.4 13.5 26.1 15.0 Transplant, n (%) 3 (5.6) 4 (9.5) 5 (7.2) 2 (7.4) mOS (95% CI), mo 15.4(13.1-30.5) 34.2(24.2-NA) 16.6(14.2-30.0) 39.8(34.2-NA) mOS, median overall survival; NA, not available; RUX, ruxolitinib; RW, real world.

Influence of peri-dose high-sensitivity troponin on clinically significant cardiac events and IL-2 interruption during post-TIL high-dose IL-2 therapy.

9557 Background: High-dose interleukin-2 (IL-2) is administered following tumor-infiltrating lymphocyte (TIL) therapy, yet cardiotoxicity frequently limits dose delivery. The clinical significance of peri-dose troponin elevations during IL-2 remains poorly defined. We evaluated troponin kinetics during IL-2 dosing and their association with significant cardiac events and IL-2 interruption. Methods: We retrospectively analyzed patients with metastatic melanoma treated with TIL therapy followed by inpatient high-dose IL-2 at Mayo Clinic Rochester (N = 24; 87 IL-2 doses). High-sensitivity troponin and ECGs were obtained before and after each IL-2 dose. Clinically significant cardiac events were defined as QTc prolongation > 480 ms, arrhythmias (excluding isolated sinus tachycardia), new ECG abnormalities, vasopressor requirement, or cardiology consultation. Dose-level outcomes included cardiac events and subsequent IL-2 dose interruption. Paired troponin changes were assessed using Wilcoxon signed-rank testing. Dose-level associations were evaluated using clustered logistic regression to account for repeated IL-2 doses within patients. Results: 24 patients received 87 IL-2 doses (median 3 [IQR 2 - 5]). Median age was 63 years (IQR 54 - 70), and 58% were male. Baseline troponin median was 11 ng/L (IQR 6 - 21); 38% had baseline troponin > 14 ng/L. Post-dose troponin exceeded 14 ng/L in 16/24 patients (67%) and 48/87 doses (52%). Median delta troponin was 0 ng/L (IQR −1 to 6; range −24 to 133), with post-dose levels higher than pre-dose values ( p = 0.02). Clinically significant cardiac events occurred in 33% of patients; no post-IL-2 reductions in left ventricular ejection fraction were observed. At the dose level, cardiac events occurred only when post-dose troponin exceeded 14 ng/L (15/48 [31%] vs 0/39 doses; p < 0.001). In clustered logistic regression, post-dose troponin ≥15 ng/L was independently associated with cardiac events (OR 9.6, 95% CI 1.5 - 60.6; p = 0.02) and subsequent IL-2 dose interruption (OR 3.4, 95% CI 1.1 - 10.7; p = 0.04). When modeled continuously, each 1-unit increase in log-transformed post-dose troponin was associated with higher odds of cardiac events (OR 3.3, 95% CI 1.6 - 6.7; p = 0.001) and IL-2 interruption (OR 2.8, 95% CI 1.6 - 4.9; p < 0.001). Baseline troponin correlated with peak post-dose troponin (ρ = 0.44, p = 0.04). Conclusions: Peri-dose troponin elevations are common during post-TIL high-dose IL-2; however, higher post-dose troponin independently identifies clinically significant cardiac toxicity and predicts IL-2 interruption after accounting for repeated dosing within patients. Baseline troponin modestly predicts peak elevations, supporting standardized, risk-stratified troponin monitoring and prospective evaluation of toxicity-adaptive IL-2 delivery strategies in TIL therapy.

Impact of lecture on knowledge of IM residents in diagnosing and managing CRS and ICANS.

e23313 Background: Cytokine release syndrome (CRS) and immune effector cell-associated neurotoxicity syndrome (ICANS) will become more prevalent in the outpatient setting as immunomodulatory drugs like chimeric antigen receptor T-cell therapy (CAR-T) and Bi-specific antibody therapy become more ubiquitous. Because of this, there is a vast need for internal medicine (IM) physicians to be aware of this. Methods: Our project entailed educational lectures for IM physicians at multiple hospitals throughout the state of Michigan in a community and academic setting. Thirty-five participants completed paired pre- and post-intervention surveys after a 1-hour educational lecture. These responses were then analyzed using a paired T-test analysis. Results: Our survey demonstrated an increase from 83% to 97% in knowledge of diagnosing CRS/ICANS and a corresponding increase from 73% to 97% in knowledge about selecting the correct treatment for CRS/ICANS. A composite toxicity confidence score (average of CRS and ICANS ratings) increased from 1.67 pre-intervention to 3.50 post-intervention, corresponding to a mean improvement of 1.83 points. Improvement was observed in nearly all participants, demonstrating a large and highly consistent effect of the educational intervention. Individually, both CRS and ICANS domains showed highly statistically significant improvements (both p < 10⁻⁹). Conclusions: Our project shows that lecture-based education on this topic in the era of emerging immune-based therapy shows robust improvement in recognition and treatment of CRS and ICANS. This can be modeled and upscaled for distribution to other disciplines, states, and institutions throughout the United States already utilizing CAR-T and bispecific antibodies.

Racial and socioeconomic disparities in stage at diagnosis and survival in pancreatic cancer: Analysis of SEER 17 data (1998–2017).

e16338 Background: Pancreatic cancer is characterized by late stage diagnosis and high mortality. While racial and socioeconomic status (SES) disparities are known to affect cancer outcomes, their specific impact on pancreatic cancer stage distribution and survival requires further quantification. This study examined racial disparities in stage at diagnosis and survival using a large, population based cohort. Methods: We analyzed 158,308 pancreatic cancer cases from the SEER 17 registries (1998–2017). Race was categorized as White, Black, Asian or Pacific Islander (API), and American Indian/Alaska Native (AI/AN). Stage was classified as Localized, Regional, or Distant. SES was determined by median household income quintiles. Survival was assessed via Kaplan Meier analysis and multivariate Cox proportional hazards models adjusting for stage and SES. Results: Among 158,308 patients, stage at diagnosis differed significantly by race (χ² = 117.56, df = 8; p < 0.001), with most patients presenting with distant-stage disease. In unadjusted analyses, overall survival differed by race (log-rank p < 0.005). In multivariable Cox proportional hazards models adjusted for stage at diagnosis and socioeconomic status, survival differed across racial groups. Using White patients as the reference group, American Indian or Alaska Native patients had a higher risk of mortality, whereas Asian or Pacific Islander patients had a lower risk of mortality. Black patients did not have a statistically significant difference in mortality compared with White patients after adjustment. Stage at diagnosis was the strongest predictor of survival, with both localized and regional disease associated with substantially lower mortality compared with distant-stage disease. Conclusions: Racial disparities in pancreatic cancer are heavily driven by stage at diagnosis, particularly for Black and AI/AN populations. The finding that Black patients’ survival did not differ significantly from the reference after adjustment for stage and SES suggests that the observed survival gap is primarily due to later diagnosis rather than biological differences. These results highlight the critical need for targeted early detection strategies in minority communities to achieve health equity.

Healthcare costs following osimertinib versus earlier-generation tyrosine kinase inhibitor initiation in non–small cell lung cancer: A U.S. claims-based study.

e20762 Background: EGFR-targeted tyrosine kinase inhibitors (EGFR-TKIs) are the standard first-line treatment for patients with EGFR-positive non-small cell lung cancer (NSCLC). Osimertinib, a third-generation EGFR-TKI, exhibited superior clinical outcomes compared to 1 st or 2 nd -generation (1/2G) TKIs; real-world data on healthcare costs after treatment initiation remain limited. The aim of this study is to compare healthcare costs following initiation of Osimertinib versus earlier-generation therapies using administrative claims data. Methods: A retrospective cohort study was conducted using the Merative MarketScan Commercial Claims and Medicare Supplemental databases (2016–2024). Chemotherapy-naïve adults with NSCLC who newly initiated Osimertinib or 1/2G-EGFR-TKIs were identified and followed for 6 months from the first date of the TKI dispensing encounter (Index Date). NSCLC status was confirmed using ICD-10 diagnosis codes within ±14 days of the Index Date. We calculated per-patient means and medians of the aggregated 6-month all-cause costs across inpatient, outpatient, and outpatient pharmacy settings. Costs were adjusted to 2025 US dollars. Results: The study included 1,355 patients from the Commercial Claims database, including 1,132 Osimertinib and 223 1/2G-TKI patients. From the Medicare Supplemental database, a total of 1,082 individuals were identified, consisting of 569 Osimertinib 513 1/2G-TKI patients. The mean age was 53 years in the Commercial cohort and 75 years in the Medicare cohort, with approximately 70% female in both populations. Mean (median) all-cause healthcare costs for Osimertinib vs. 1/2G-TKI from the Commercial claims cohort were $169,937 ($147,536) vs. $102,221 ($89,156), respectively, consisting of the inpatient costs of $12,710 vs. $13,415, outpatient care costs of $35,620 vs. $23,789, and pharmacy costs of $121,607 vs. $65,017. From the Medicare Cohort, the total respective costs of $139,303 ($134,147) vs. $99,575 ($83,954) comprised $10,265 vs. $8,631 inpatient medical costs, $21,917 vs $29,952 outpatient medical costs, and $107,122 vs. $60,992 pharmacy costs. Conclusions: Treatment with Osimertinib was associated with higher total healthcare costs compared with 1/2G-TKIs, with consistent findings across both Commercial and Medicare cohorts. The cost escalation driven by higher outpatient pharmacy costs might not be fully offset by cost savings from the medical sectors, which should be further explored by additional research considering changes in treatment patterns and adjustments for patient characteristics.

Use of G8 screening and CARG toxicity scores to predict emergency room (ER), hospital (IP) admissions, and mortality in newly diagnosed older patients with cancer undergoing chemotherapy with or without immunotherapy: Updated 3-year analysis.

e13768 Background: National guidelines recommend that older patients undergo geriatric screening, G8 and CARG. In our initial six-month pilot at Northern California oncology clinics, high risk (G8 ≤14) and CARG (10-19) scores correlated with higher mortality and ER/IP admission in older adults with newly diagnosed cancer. We then extended this workflow to all Kaiser Northern California clinics. Methods: Patients 65 and above with new cancer diagnoses (Jan 2021–Dec 2024) completed G8 and CARG assessments. Monitored outcomes were ER/IP admissions and mortality. Associations between risk categories and demographics/utilization were tested using chi-square, while Cox models analyzed relationships between scores and time to ER/hospitalization or death. Results: A total of 2,684 patients completed G8, and 4,347 patients completed CARG. The proportion of patients classified as high risk by G8 or CARG increased with each decade of age (G8: < 70 years, 64%; 70–79 years, 65%; 80–89 years, 87%; ≥90 years, 100%; p < 0.0001; CARG: < 70 years, 13%; 70–79 years, 33%; 80–89 years, 53%; ≥90 years, 71%; p < 0.0001). High-risk CARG was more prevalent among men compared to women (33% vs. 29%, p < 0.0001). There was no significant association between ethnicity and high risk G8 or CARG. High risk CARG scores were noted in upper GI (42%), genitourinary (GU) (42%), and lower GI (43%) cancers patients. High risk G8 scores were prevalent in upper GI (85%), thoracic (70%), malignant hematologic (69%), and lower GI (66%) cancers. Both assessment tools demonstrated statistically significant differences among cancer types (p < 0.0001). The rates of mortality, ER/IP admission, and referral to hospice or palliative care increased at 180 days in both high risk G8 and CARG groups (p < 0.0001). Patients with high risk G8 or CARG scores had 3x the mortality rate, 1.4x the ER/IP admission rate, and over 2x as likely to be referred to palliative care or hospice compared to low risk patients. Conclusions: Our follow up analysis with an expanded sample size corroborates previous results, demonstrating that high risk scores on G8 and CARG assessments at initial cancer diagnosis are predictive of ER/IP admissions and mortality in older adults with cancer. These assessments should be incorporated into the initial oncology evaluation to inform shared treatment decision making. Hazard ratios for ER/IP admission and mortality comparing high/low risk G8 and CARG patients. Outcome HR 95% CI p-value High vs Low Risk CARG ER/IP 1.409 1.231- 1.613 <.0001 High vs Low Risk G8 ER/IP 1.332 1.201- 1.478 <.0001 High vs Low Risk CARG Mortality 2.208 1.761-2.769 <.0001 High vs Low Risk G8 Mortality 1.894 1.594-2.25 <.0001 Age, cancer type, and BMI also showed significance in determining outcomes.

A phase 1 study to evaluate safety, tolerability, and pharmacokinetics of LNP8701 (SOS1 inhibitor) in subjects with metastatic solid tumors.

e15162 Background: SOS1 (Son of Sevenless1) is a critical regulator of RAS signaling, facilitating RAS activation to drive cell proliferation and survival. Dysregulated SOS1 activity contributes to oncogenic signaling and tumor progression. LNP8701 is an orally active, investigational SOS1 inhibitor designed to block SOS1-mediated RAS activation, thereby limiting oncogenic signaling that promotes tumor growth. Here, we present results from the dose-escalation (DE) sub-study from a Phase 1 study of LNP8701 in patients with metastatic solid tumors. Methods: This is a first-in-human, multicenter, open-label study in patients with metastatic solid tumors with failed prior standard therapies or for whom no standard therapy exists. DE followed initial modified acceleration followed by 3+3 design based on DSMB recommendation. LNP8701 was administered orally once daily for 6 cycles of 28 days each. Primary objective was to determine maximum tolerated dose (MTD) assessed by dose-limiting toxicities (DLT) in cycle 1; secondary objectives included safety, pharmacokinetics and preliminary antitumor activity measured by RECIST v1.1. Results: 20 patients (10 males & 10 females) of stage IV cancer including 5 patients of cervical cancer, 3 patients each of kidney and lung cancer, 2 patients of rectal cancer and 1 patient each of vaginal vault, colon, colorectal, pancreatic, ovarian, breast cancer and left thigh sarcoma were enrolled in DE study. Patients received LNP8701 at doses of 100 mg (n = 1), 200 mg (n = 7), 300 mg (n = 7), 400 mg (n = 5). Median age was 50.5 (range 29 - 70) years. A total of 16 (80%) patients had at least 1 treatment emergent adverse event (TEAE). The majority (80%) of TEAEs were unrelated to LNP8701. The most common TEAEs were anemia (40%), vomiting (25%), and pain in extremity (25%). A total of 6 (30%) patients had serious TEAEs. One patient in 200 mg cohort had a DLT of Grade 4 thrombocytopenia during cycle 1. None of the patients in 100 mg, 300 mg and 400 mg cohort had DLT in cycle 1. Fifteen patients discontinued treatment due to progressive disease (n = 14) and SAE (n = 1) as of data cut-off date 21-Jan-2025. Three patients continued LNP8701 beyond 6 cycles on compassionate basis, of which 2 patients (1 patient each of Ca vaginal vault and Ca breast) completed 12 cycles with stable disease and 1 patient with Ca cervix completed 14 cycles with stable disease. The patient with Ca vaginal vault had partial response at the end of cycle 4 and cycle 6. Overall, 5 (25%) patients had stable disease and 1 (5%) patient had partial response as the best overall response during the study. After single and multiple dosing, peak plasma concentrations of LNP8701 were observed between 30 mins to 1 hr. Conclusions: LNP8701 was well tolerated with desirable safety, PK and preliminary efficacy profile. Clinical trial information: CTRI/2024/08/072373.

Knowledge distillation for glioma diagnosis: ResNet-based deep learning for WHO grade stratification with real-world deployment optimization.

2083 Background: Gliomas are the most common primary CNS malignancies. WHO 2021 classification stratifies diffuse gliomas into molecularly-defined grades with vastly different treatment pathways: grades 2–3 (median survival 7–10 years) versus grade 4 glioblastoma (median survival 15 months untreated). Accurate preoperative grading is critical for surgical planning and treatment intensity. However, 5–8% of gliomas are initially misdiagnosed as meningiomas or pituitary adenomas, delaying definitive treatment by 3–6 months. Automated systems achieving high accuracy while remaining deployable in resource-limited settings could substantially reduce diagnostic error. Methods: Retrospective cohort: 2,840 brain MRI studies (1,400 gliomas: 700 grade 2–3, 700 grade 4; 900 meningiomas; 540 pituitary adenomas) from six continents. Ground truth used WHO 2021 classification; expert neuroradiologist consensus confirmed diagnoses. ResNet152 (60.2M parameters; 224×224 input; 28.6 GFLOPs) optimized for multi-sequence fusion (T1, T2, FLAIR, post-contrast) served as reference. ResNet18 (11.7M parameters; 224×224 input; 1.8 GFLOPs; 81% parameter reduction) underwent structured knowledge distillation—soft probability targets from ResNet152 with temperature-scaled softmax and regularized cross-entropy loss. Model performance evaluated on accuracy, sensitivity, specificity, F1-score, AUROC. Differential diagnostic accuracy assessed. Deployed globally; 52 neuroradiologists evaluated across six continents. Results: Glioma grading: ResNet18 achieved 96.9% accuracy for grade 2–3 vs. grade 4 (sensitivity 97.8%, specificity 96.0%, AUROC 0.991). Differential diagnosis: 95.1% for glioma vs. meningioma; 96.0% for glioma vs. pituitary. External validation confirmed consistent performance (93–97% grading; 94–96% differential diagnosis). Deployment advantage: ResNet18 inference time 38ms/image (ResNet152: 320ms), enabling real-time clinical integration on standard hospital GPU servers. Global deployment: 94.6% of neuroradiologists rated the system clinically valuable. Knowledge distillation preserved 99.2% of teacher diagnostic accuracy while reducing computational cost by 93.7%. Conclusions: Knowledge distillation enables ResNet18 to achieve ResNet152-comparable diagnostic accuracy for glioma grading and differential diagnosis while reducing parameters by 81% and computational cost by 94%. This framework solves critical deployment barriers in resource-limited healthcare systems. ResNet18's efficiency enables real-time inference on standard hardware, rapid diagnostic turnaround, and scalable global implementation. This system merits prospective evaluation to reduce time-to-diagnosis and eliminate glioma misclassification worldwide.

An exploratory study to assess the safety, immunogenicity, and preliminary anti-tumor activity of the EBV mRNA vaccine (WGc-043 injection) combined with PD-1 antibody in patients with relapsed or refractory ENKTL.

7077 Background: Extranodal NK/T cell lymphoma (ENKTL) is a rare and aggressive lymphoma associated with Epstein-Barr virus (EBV) infection and requires novel therapeutic strategies. WGc-043, a therapeutic mRNA vaccine, showed promise against ENKTL patients alone by targeting EBV-driven oncogenesis in our previous study. This study evaluates the safety, immunogenicity, and preliminary anti-tumor activity of WGc-043 combined with PD-1 antibody in patients with relapsed or refractory ENKTL (r/r ENKTL). Methods: This single-center, open-label, exploratory study enrolled 5 patients with EBER1/2 positive r/r ENKTL. The primary inclusion criteria were patients with an ECOG of 0-2 who have failed an asparaginase-based regimen. WGc-043 was administered intramuscularly, including 5 doses of base immunization followed by optional personalized treatment. The base immunization was given at week 1, 2, 3, 4 and 7. Tislelizumab (PD-1 antibody) was administered every three weeks until disease progression, for a maximum of two years. The primary endpoint is safety. Secondary endpoints include immunogenicity, ORR and DCR. The exploratory analysis focuses on immune biomarkers predictive of therapeutic response. Results: WGc-043 combined with PD-1 antibody (Tislelizumab) demonstrated a favorable safety and efficacy in patients with r/r ENKTL. It was well tolerated with only grade 1 adverse events. The most common adverse events were injection site reactions (100%, 4/4) , hemoglobin decreased (50%, 2/4), TSH elevated (50%, 2/4), fever (25%, 1/4), weakness (25%, 1/4) and lymphopenia (25%, 1/4). Importantly, this combo showed good, durable responses, with 2 patients achieving CR in 4 evaluable patients, and both remained CR status on the Jan 11, 2026. This result indicates that the combination of WGc-043 and PD-1 antibody is promising as part of the therapeutic strategy for this challenging lymphoma. Conclusions: WGc-043 combined with PD-1 antibody is well tolerated with a safety profile and yields encouraging CR rate and durable responses. Further investigations are warranted to validate these findings and to explore the full potential of this regimen in larger studies. Clinical trial information: 2200065962. Efficacy and safety. Response category Number of subjects CR 2 PR 0 SD 0 PD 2 Total 4 ORR 50% DCR 50% Adverse events Grade 1 Grade 2 Grade 3 Grade 4 Grade 5 Any adverse event 4/4 (100%) 0 0 0 0 Injection site reaction 4/4 (100%) 0 0 0 0 Hemoglobin decreased 2/4 (50%) 0 0 0 0 TSH elevated 2/4 (50%) 0 0 0 0 Fever 1/4 (25%) 0 0 0 0 Weakness 1/4 (25%) 0 0 0 0 Lymphopenia 1/4 (25%) 0 0 0 0

Copy number aberration profiling using shallow whole-genome sequencing-based LeukoPrint in multiple myeloma.

7554 Background: Multiple myeloma (MM) is a clonal plasma cell malignancy in which cytogenetic abnormalities are central to prognosis and therapy. Karyotyping is limited by low plasma cell proliferation, and FISH provides only targeted, costly detection. LeukoPrint, a shallow whole-genome sequencing–based assay, enables genome-wide CNA profiling. Supported by prior validation, we launched a prospective multicenter trial (ChiCTR2300077768) to assess the clinical relevance of LeukoPrint-detected CNAs in MM, and report interim results from the first 102 patients. Methods: This multicenter, prospective trial aims to enroll 396 newly diagnosed active multiple myeloma patients per IMWG 2014 criteria. As of June 12, 2025, 102 patients from seven hospitals in China were enrolled. LeukoPrint was used to profile CNAs (>1 Mb) from paired bone marrow CD138⁺ plasma cell gDNA and peripheral blood cfDNA collected at diagnosis, post-induction, and relapse. Karyotyping and FISH were performed in parallel when available and compared with LeukoPrint. Risk stratification followed mSMART criteria. Results: LeukoPrint identified 863 CNAs (379 chromosomal and 484 sub-chromosomal) in 91 of 102 patients (89.2%). Hyperdiploidy was observed in 33 patients, non-hyperdiploidy in 58, and normal karyotype in 11. The most frequent alterations were 1q gain (53%) and chromosome 13 deletion (50%). Odd-numbered trisomies, often co-occurring, were detected in 51.0% of patients, predominantly in hyperdiploid cases, whereas del(13) was evenly distributed across ploidy groups. Subdiploidy was common (54%) and hyperdiploidy rare (13%) among IGH-translocated cases, suggesting mutual exclusivity of hyperdiploidy and IGH-translocation. LeukoPrint showed 88.4% concordance with FISH for four key prognostic CNAs (1p32-, 1q21+, 13q14-, 17p13-) but detected substantially more abnormalities, including in 80% of FISH-negative patients. It also markedly outperformed karyotyping (90.1% vs. 8.8%), consistent with the limited sensitivity of karyotyping due to the low proliferative activity of plasma cells in MM. Blood-based ctDNA analysis showed good concordance with bone marrow results (81.9%), and CNA dynamics tracked closely with clinical remission, supporting LeukoPrint as a robust tool for diagnosis and treatment monitoring in MM. Conclusions: LeukoPrint outperforms FISH and karyotyping, providing a cost-effective, genome-wide CNA platform in MM and demonstrating utility in ctDNA-based liquid biopsy. These results support LeukoPrint as a complementary, and potentially alternative approach to conventional cytogenetics. Clinical trial information: ChiCTR2300077768.

Proteome analysis refines molecular processes underlying metamorphosis in the ascidian Ciona intestinalis

Tunicates, including ascidians, are recognized as the true ‘sister group’ of vertebrates and are emerging as models to study larval and post-larval development, including degeneration of central nervous system (CNS), in chordates. Ascidian larvae have the typical chordate body plan that includes a dorsal neural tube. During their metamorphosis, a deep tissue reorganization takes place, with some tissues that degenerate while others develop to become functional during the adult life. The larval CNS also degenerates, and most neurons disappear, making room for the formation of adult CNS. The genome of the ascidian Ciona intestinalis has been sequenced and annotated, with several CNS specific genes that have been characterized. These features make ascidian metamorphosis a good model to study the mechanisms underlying physiological CNS degeneration. To shed light on the molecular determinants of C. intestinalis metamorphosis, we analyzed the proteome at three stages of development: swimming larva (SwL, Hotta stage 28), settled larva (SetL, Hotta stage 32) and metamorphosing larva (MetL, Hotta stage 34). A total of 405 modulated proteins were identified by mass spectrometry by comparing the three stages. Enrichment and network analysis showed the involvement of several processes/pathways, including autophagy and mTOR pathways, and actin cytoskeleton organization and remodeling among the most significant ones. This study helps to elucidate the molecular processes underlying ascidian metamorphosis and provides insight into mechanisms of physiological neurodegeneration in ascidians.

Rational design of multifunctional nanostructured Ag-Doped ZnO-TiO2 derivatives for effective photocatalytic and antioxidant mechanisms

Large‐Area 2D Metasurface‐Based Triboelectric E‐Skin Arrays: Contact & Proximity Tactile Mapping with Broadband Acoustic Readouts (Adv. Mater. 36/2026)

Erratum to “Successive Orthorhombic Distortions in Kagome Metals by Molecular Orbital Formation”

Co-overexpression of ELOVL2 and ELOVL5 promotes clear cell renal cell carcinoma progression through LIMK1-mediated cytoskeletal reorganization

Structural basis of the protein kinase PKN1 HR1 domain oligomerization and differential regulation by RhoA and Rac1

Updated results of surufatinib combined with gemcitabine and cisplatin and immune checkpoint inhibitor (ICI) for unresectable locally advanced or metastatic intrahepatic cholangiocarcinoma.

4136 Background: Advanced metastatic intrahepatic cholangiocarcinoma (ICC) was characterized by poor survival and limited therapeutic options. Surufatinib, a selective tyrosine kinase inhibitor targeting VEGFR 1, 2, and 3, FGFR1, and CSF-1R, provides dual anti-tumor action via anti-angiogenesis and tumor microenvironment regulation. Surufatinib combined with immunotherapy and chemotherapy may enhances anti-cancer effects for advanced metastatic ICC. This study evaluates the effectiveness and safety of surufatinib combined with gemcitabine, cisplatin, and immune checkpoint inhibitor (ICI) as the first-line treatment for patients with unresectable locally advanced or metastatic intrahepatic cholangiocarcinoma. Methods: This is an open-label, single-arm, single-center trial. Eligible patients who were 18 -75 years old with histologically confirmed unresectable locally advanced or metastatic ICC were enrolled. Pts received surufatinib (250mg, orally, once daily), ICI (Zimberelimab or Toripalimab, 240mg, intravenous infusion, d1, q3w), and chemotherapy (gemcitabine 1000mg/m 2 intravenous infusion, 30min, d1, d8, q3w; cisplatin 25mg/m 2 intravenous infusion, 2h, d1, d8, q3w) until disease progression, death, surgery, intolerable toxicity, or withdrawal of consent. The primary endpoint was objective response rate (ORR). Secondary endpoints included progression-free survival (PFS), disease control rate (DCR), overall survival (OS), conversion to surgical resection rate, pathological complete response rate (pCR), and safety. Results: By January 20, 2026, 20 pts were enrolled with median age 60 years (range: 43-70), male 50%, 100% ECOG PS 0-1, 95% stage IV. 20 pts were efficacy evaluable, 5% (1/20) pts achieved complete response (CR), 50% (10/20) pts achieved partial response (PR), 35% (7/20) pts achieved stable disease (SD). The confirmed ORR was 55%, DCR was 90%. The median PFS was 7.7m (95%Cl: 6.3-11.3 months), and the median OS was 17.8m (95%Cl: 16.1-21.8 months). The most common TEAEs of all grades were hypertension (40%), edema (15%) and vomiting (10%), 1 patient was Immune-related pneumonia occurred. No grade ≥ 3 TEAEs or new safety signals occurred. Conclusions: Surufatinib plus gemcitabine and cisplatin combined with immune checkpoint inhibitor, the chemotherapy regimen as the standard treatment in combination with targeted and immunotherapy showed preliminary anti-tumor activity and manageable toxicity for the 1L treatment of ICC, providing an additional treatment option for pts with ICC. This study has been completed and the paper has been prepared for submission, the final results details will be published after the conference. Clinical trial information: ChiCTR2400085526.