Citations

Abstract Background: Early-onset gastrointestinal (GI) cancers are increasing globally, yet the genomic basis for age-related differences across multiple tumor types remains poorly defined. We leveraged AACR Project GENIE v18.0 (N≈34,000 GI tumors) to characterize molecular distinctions by age in colorectal, pancreatic, small bowel, and anal cancers with sufficient early-onset representation. Methods: Clinical and genomic data were extracted and tumors classified into early-onset (<50 years) and late-onset (≥50 years). Mutation frequencies were compared by cancer type and age group, focusing on recurrent drivers, DNA repair, Wnt signaling, RAS/MAPK, chromatin remodeling genes, and targetable alterations. Results: Early-onset colorectal cancer (N=5,724) showed increased Wnt/β-catenin pathway alterations: APC (+4.9%), TCF7L2 (+4.1%), and TP53 (+5.0%), while late-onset tumors (N=15,807) were enriched for BRAF (−5.4%), RNF43 (−1.9%), and GNAS (−1.1%), indicating divergent tumorigenic pathways. Early-onset pancreatic cancer (N=890) exhibited significantly lower frequencies of canonical drivers—KRAS (−24.7%), TP53 (−15.6%), CDKN2A (−6.9%), SMAD4 (−6.0%)—with enrichment of MEN1 (+3.5%), TSC2 (+3.0%), ATRX (+2.0%), DAXX (+1.9%), and CTNNB1 (+2.4%), consistent with fusion-driven, hereditary, and chromatin remodeling biology. Small bowel cancer (N=99) in early-onset patients demonstrated enrichment in DNA damage response and MAPK/RAS genes including BRCA2 (+7.2%), NF1 (+7.4%), FBXW7 (+8.8%), FLT4 (+6.5%), ARAF (+6.4%), and ALK (+5.1%), whereas late-onset tumors (N=538) were enriched for KDR (−5.1%) and MGA (−5.2%). Early-onset anal cancer (N=62) showed elevated MGA (+7.8%), WHSC1/NSD2 (+6.9%), PBRM1 (+6.2%), FANCA (+5.2%), ERCC5 (+6.2%), and MSH6 (+5.9%), highlighting chromatin remodeling and DNA repair pathways potentially related to HPV biology; late-onset tumors (N=464) had enrichment in ATM (−5.6%), PTEN (−5.5%), STK11 (−5.2%), APC (−5.2%), and KMT2C (−5.0%), reflecting age-associated genomic instability. Conclusions: This pan-GI age-stratified analysis reveals distinctive oncogenic programs in early-onset GI cancers characterized by Wnt/TP53-driven colorectal, fusion- and chromatin remodeling-driven pancreatic, DNA damage response/MAPK-driven small bowel, and chromatin-immune DDR signatures in anal cancers. These findings underscore the need for age-tailored precision oncology and support routine genomic profiling in early-onset GI malignancies to identify targetable alterations enriched in younger patients. Citation: AACR Project GENIE Consortium. AACR Project GENIE: Powering Precision Medicine Through an International Consortium. Cancer Discov. 2017;7(8):818-831. (Data: GENIE Release 18.0-public)Acknowledgement: AACR and Project GENIE registry Citation Format: Ahmed Abdelhakeem, Nency Ganatra, Dina Elantably, Tayo Adeoye, Nayef H. Abdel-Razeq, Mosalem M. Osama, Hani M. Babiker. Distinct age associated genomic landscapes in colorectal, pancreatic, small bowel, and anal cancers: A pan gastrointestinal analysis of 34,000 tumors from AACR Project GENIE v18.0 [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 4105.

Abstract Background: PARP inhibitors are restricted to a subset of BRCA-associated cancers, but homologous recombination repair (HRR) deficiency signifies a broader genomic vulnerability that may extend therapeutic opportunities across diverse malignancies. We utilized AACR Project GENIE v18.0-public to define the pan-cancer prevalence of HRR gene alterations. Methods: Somatic mutation and clinical sample data from GENIE v18.0-public were analyzed. HRR genes included BRCA1, BRCA2, ATM, CHEK2, PALB2, RAD51B/C/D, RAD51L1, BARD1, BRIP1, and core FANC genes. HRR alteration rates were calculated by tumor type using unique samples; only de-identified aggregate results are reported. Results: Among 101,240 tumors, HRR gene alterations were frequent in multiple non-BRCA-associated cancers. Highest rates were seen in sialoblastoma (100%), large cell medulloblastoma (50%), malignant glomus tumor (50%), lung cancer (46.0%), infantile fibrosarcoma (42.9%), liver (40%), kidney (40%), non-melanoma skin (36.9%), colorectal (36.6%), melanocytoma (34.2%), biliary tract (33.3%), bladder (31.5%), and ovarian/fallopian tube (30.2%). Notably, lung, bladder, hepatobiliary, renal, gastrointestinal, and some sarcoma subtypes showed HRR mutation prevalence equal or higher than current PARP indications, yet lack FDA approval. Conclusions: HRR alterations are widespread across solid tumors, often surpassing rates in cancers eligible for PARP treatment. These findings support expanding tumor-agnostic HRR genomic testing beyond BRCA-linked malignancies and warrant prospective trials of PARP inhibitors in HRR-enriched tumor types such as lung, bladder, hepatobiliary, renal, gastrointestinal, and pediatric/rare cancers. Impact: Large-scale genomic profiling from AACR Project GENIE demonstrates that HRR gene alterations are common across diverse solid tumors, with prevalence in multiple cancer types surpassing those currently eligible for PARP inhibitor therapy. These findings support broad implementation of HRR genomic testing in routine molecular diagnostics and provide a rationale for expanding PARP inhibitor use beyond BRCA-associated malignancies. The data justify prospective tumor-agnostic clinical trials of PARP inhibitors in HRR-enriched cancers such as lung, bladder, hepatobiliary, renal, gastrointestinal, and select pediatric and rare tumors; this could transform therapeutic decision-making and accelerate development of synthetic lethality-based targeted therapies Citation: AACR Project GENIE Consortium. Cancer Discov. 2017;7(8):818-831 Acknowledgement: AACR and Project GENIE Citation Format: Ahmed Abdelhakeem, Nency Ganatra, Dina Elantably, Tayo Adeoye, Nayef H. Abdel-Razeq, Osama M. Mosalem, Hani M. Babiker. Pan cancer enrichment of homologous recombination repair (HRR) gene alterations in 101,240 tumor samples from AACR Project GENIE: Implications for tumor-agnostic PARP inhibitor expansion [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 4106.

Abstract Introduction: Diffuse large B-cell lymphoma (DLBCL) is the most common type of malignant lymphoma. Previous studies have shown that DLBCL cells are susceptible to GPX4 (Glutathione peroxidase 4)-regulated ferroptosis, an iron-dependent type of programmed cell death characterized by increased reactive oxygen species and excessive lipid peroxidation. In addition to GPX4, FSP1 (Ferroptosis suppressor protein 1, previously known as AIFM2) is a glutathione-independent repressor of ferroptosis that has shown prognostic significance in solid tumors. In this study, we aimed to reveal the significance of GPX4 and FSP1 in DLBCL. Patients and Methods: We performed immunohistochemistry (IHC) for GPX4 and FSP1 in a large cohorts of patients with de novo DLBCL, and evaluated their cytoplasmic and nuclear expression. Prognostic analysis was performed for FSP1 and GPX4 expression in patients treated with rituximab (R)-CHOP or CHOP chemotherapy, respective of TP53 mutation status, as p53 regulates ferroptosis and TP53 mutation is associated with poorer prognosis in DLBCL. Results: FSP1 expression with a ≥10% cutoff was positive in 42.3% of DLBCL cases. Patients with TP53 mutations had a non-significant trend of higher mean FSP1 expression than those with wild-type (Wt) TP53 (P=0.11). Prognostic analysis revealed that cytoplasmic FSP1+ expression was associated with significantly poorer overall survival and progression-free survival in Wt-TP53 DLBCLs treated with R-CHOP (P=0.016 and P=0.003, respectively), and only showed non-significant unfavorable trends in TP53 mutated DLBCLs and patients treated with CHOP. In contrast, GPX4 expression did not show a significant unfavorable prognostic effect in DLBCL, and in fact, was associated with a non-significant trend of better survival. Previously we have quantified immune cell abundance and PD-1/PD-L1 expression in the tumor microenvironment of the study cohort using multiplex fluorescent IHC. Correlative analysis found that FSP1+ patients had significantly higher mean and median abundance of CD68+ cells (both M1 and M2 macrophages) and CD11c+ cells than FSP1- patients. Further analysis in Wt-TP53 and mutated TP53 subcohorts found that only in the TP53 mutated DLBCL subcohort, FSP1 expression was associated with significantly higher mean and median CD163-CD68+ (M1) and overall CD68+ macrophages, whereas in the Wt-TP53 DLBCL subcohort, FSP1 expression was associated with significantly higher median (but not mean) CD163+CD68+ (M2) macrophages and CD11c+ cells. Summary: Cytoplasmic FSP1 expression but not GPX4 had significantly adverse prognostic effect in patients with Wt-TP53 DLBCL treated with standard immunochemotherapy. Our results also suggest that in addition to ferroptosis regulation, macrophage abundance was relevant for the prognostic effects of FSP1 expression. Citation Format: Beibei Lyu, Zijun Xu-Monette, Xiaoxian X. Zhao, Eric D. Hsi, Ming Chen, Carlo Visco, Alexandar Tzankov, Karen Dybkaer, Mu-En Wang, Chang Wang, Qingyan Au, Harry Nunns, Zenggang Pan, Benjamin Parsons, Santiago Montes-Moreno, Fenghuang Zhan, Michael B. Møller, Leon Bernal-Mizrachi, Youli Zu, Shanxiang Zhang, Weina Chen, Govind Bhagat, Yong Li, KEN H. YOUNG. Expression of the ferroptosis suppressor FSP1 but not GPX4 shows significant adverse prognostic effect in diffuse large B-cell lymphoma with wild-type TP53 [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 4661.

Abstract Genome-wide association studies (GWAS) have identified thousands of genetic variants associated with individual cancer risk, and polygenic risk scores (PRS) are typically derived from these cancer-specific variants. Although previous pan-cancer studies revealed shared genetic architectures across cancer types, latent factors contributing to cancer risk have not been explored. This may allow for improved prediction of cancer risk based on a latent cancer factor which leverages shared variants across cancers. Here, we conducted an integrative pan-cancer GWAS across seven solid tumors (breast, ovarian, endometrial, pancreatic, lung, colorectal, melanoma) to identify shared latent genetic factors and novel pleiotropic variants. We curated the most recent and large-scale GWAS summary statistics to maximize statistical power, comprising 398,917 cases and 1,501,715 controls of European ancestry. Shared genetic structures among 6,346,960 common variants were evaluated using Genomic Structural Equation Modeling (GenomicSEM). Three latent factors (Common Cancer Factor, Female-Specific Cancer Factor and Non-Sex Specific Cancer Factor) were constructed, and associations of individual variants with each latent factor were estimated. Genome-wide significant loci (p < 5x10-8) were functionally annotated and subjected to pathway enrichment analysis. Models capturing shared genetic risk across cancers fit the data well. A single “Common Cancer Factor” explained much of the shared risk (SRMR = 0.064 [lower is better], CFI = 0.938 [closer to 1 is better]), and a correlated two-factor model—distinguishing female-specific from other cancers—showed even stronger fit and biological relevance (SRMR = 0.055, CFI = 0.988). The one-factor (Common Cancer Factor; SRMR: 0.064, CFI: 0.938, AIC: 54.7, p_chisq: 0.021) and the two-factor (Female-Specific Cancer Factor and Non-Sex Specific Cancer Factor; SRMR: 0.055, CFI: 0.988, AIC: 45.4, p_chisq: 0.281) models demonstrated robust model fit and biological relevance. Factor-specific GWAS identified 233 genome-wide significant loci (p < 5x10-8) with 24 distinct loci showing shared susceptibility across cancers. More importantly, we identified eight novel variants that were not previously reported or related with any cancer risk, potentially regulating cancer-related genes such as PRC1, CEBPB, SMC2, KLF5 and ATXN2. Further analyses are needed to elucidate their roles in tumorigenesis. Our findings reveal a shared genetic architecture across major solid tumors and uncover novel pleiotropic variants with potential regulatory roles in oncogenesis. This study provides novel insights into the genetic basis of cancer susceptibility and supports the development of multi-cancer PRS to improve prediction in cancer prevention and prognosis. Citation Format: Xunxuan Chen, Gamaliel T. Taengwa, Brandon J. Coombes, Stacey J. Winham. Pan-cancer genome-wide study reveals shared genetic architecture and novel pleiotropic variants across seven solid cancers [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 3600.

Abstract Background: Liquid biopsy approaches for cancer detection often rely on tissue-specific methylation or fragmentation features to infer tumor origin. However, a pan-aerodigestive cancer detection strategy is clinically valuable given shared risk exposures, field cancerization, and frequent diagnostic ambiguity among lung, esophageal, and head & neck cancers. We hypothesized that advanced-stage disease exhibits convergent chromatin remodeling, producing a dominant malignancy-associated fragmentomic signal that supersedes subtler tissue-of-origin patterns. Methods: Whole-genome EM-seq (∼15×) was performed on plasma cfDNA from 120 patients with stage III/IV aerodigestive cancers (lung, esophageal, head & neck; squamous and adenocarcinoma) and 60 age-matched controls. Four fragmentomic modalities were assessed: (1) genome-wide fragmentation in 5-Mb bins (DELFI-like), (2) NMF spectral components, (3) 4-mer end motifs, and (4) promoter-centric nucleosome scores (Griffin). All preprocessing occurred strictly within each training fold of a repeated stratified 10-fold CV (100 repeats) to prevent data leakage. No test-sample information was used in training. Hyperparameters were tuned only within training partitions. Random Forest classifiers were trained under this fully nested scheme. Results: The multimodal classifier robustly detected advanced aerodigestive malignancies with a mean AUROC 0.86 and AUPRC 0.94. However, modality dissection revealed a clear performance hierarchy: nucleosome positioning (Griffin) alone achieved AUROC 0.85 and AUPRC 0.93, nearly matching the full integrated model, whereas end motifs performed poorly (AUROC 0.61). Biological interpretation demonstrated prominent nucleosome depletion centered at promoters of proliferation-associated genes, indicating a convergent epigenetic state of promoter opening across late-stage aerodigestive malignancies. Conclusions: Advanced aerodigestive cancers exhibit a dominant, convergent chromatin remodeling signature detectable in plasma cfDNA, driven primarily by altered nucleosome positioning. In late-stage disease, this shared malignancy-associated fragmentomic signal outweighs tissue-of-origin-dependent patterns. These findings highlight the utility of nucleosome-based fragmentomics not merely for detection, but for resolving diagnostic ambiguity in patients with indeterminate masses and for monitoring total tumor burden in high-risk aerodigestive malignancies. Citation Format: Axel Misael Hidalgo, Ayesha Hashmi, Jeff Szymanski, Pradeep Chauhan, Lilli Greiner, Faridi Qaium, Daniel J. Ma, David M. Routman, Katie M. Van Abel, Aadel A. Chaudhuri.. Convergent chromatin remodeling enables pan-aerodigestive detection of advanced malignancies in plasma cfDNA [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 3862.

Abstract Background: In metastatic cancer, Circulating Tumor cells (CTCs) are established prognostic indictors of patients (pts) less likely to respond to new lines of systemic therapy, with poor clinical outcomes, such as shorter progression free survival (PFS) and overall survival (OS). However, CTCs are typically found in specific malignancies (breast, prostate & colon), often in <20% of pts with metastatic disease, and pts without CTCs may also rapidly progress. Recently, an inflammatory pro-tumorigenic macrophage emanating from tumor stroma (i.e. Cancer associated macrophage-like cell [CAML]) was found in >90% of metastatic cancer pts, and whose phagocytic engorgement appears to correlate with poor outcomes, independent of CTCs. As CTCs and CAMLs are isolated in conjunction from a single blood sample, and both are prognostic for outcomes, we evaluated their utilization prior to induction of new systemic therapy in 6 types of metastatic cancer to model pt risk stratification based on 2 year outcomes. Methods: A prospective 2 year blind multi-institutional study was undertaken to model CTCs and CAMLs in prognosticating outcomes prior to induction of a new line of systemic therapy (n=233) in metastatic: Breast (n=60), Prostate (n=40), Pancreas (n=25), Colon (n=28), Renal Cell Carcinoma (RCC) (n=39), and Lung (n=40). Blood was filtered by CellSieveTM filters with subtypes of CTCs & hyper-enlarged CAMLs (≥100µm) enumerated. A machine learning algorithm was trained on this initial data set to develop predictive models which could stratify pt populations by risk for likelihood of PFS & OS over 2 years, including known clinical variables. Results: CTCs were absent in 80% (n=185/233) of pts, and their absence was prognostic for better PFS (HR=1.6, p=0.046), but not OS (HR=1.3, p=0.2045). In parallel, enlarged CAMLs (≥100µm) were found in 28% (n=51/185) of pts without CTCs, and were also prognostic for worse PFS (HR=2.0, p=0.0082) and OS (HR=1.9, p=0.0412). Specifically, CTCs were found in 55% (n=33) breast, 20% (n=8) prostate, 12% (n=3) pancreas, 11% (n=3) colon, 0% (n=0) RCC, and 0% (n=0) lung pts. Enlarged CAMLs were found in 45% (n=27) breast, 20% (n=8) prostate, 44% (n=11) pancreas, 44% (n=17) colon, 18% (n=7) RCC and 20% (n=8) lung pts. Overall, models indicated that ≥1 CTC (n=47) had mPFS=3.9 & mOS=14.9, while pts with 0 CTCs and ≥100µm CAMLs (n=51) had a mPFS=6.9 & mOS=13.8, and pts with 0 CTCs and <100µm CAMLs (n=134) had a mPFS=10.7 & mOS>24 months. Conclusions: These initial models confirm that both CTCs and enlarged CAMLs are prognostic indicators of worse PFS & OS. The simultaneous quantification both CTCs and CAMLs allows for more accurate pan-cancer risk stratification in an array of cancer pt populations. Additional clinical variables incorporated into the models may allow better risk subtyping and possibly forecast optimal treatment regimens. Citation Format: Daniel L. Adams, Steven H. Lin, Massimo Cristofanilli, Carolina Reduzzi, Saranya Chumsri, Martin J. Edelman, Susan Tsai, Raymond C. Bergan, Mohammed Aldakkak, Thai H. Ho, Cha-Mei Tang. Combining circulating tumor cells and cancer associated macrophage-like cells enhances risk stratification models in pan-cancer metastatic disease [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 1068.

Background: Large language models (LLMs) are being used by surgeons for education and reference yet concerns about hallucinations and reliability limit safe adoption. Retrieval-augmented generation (RAG) can offer a potential solution by grounding responses in a high-quality external database (e.g., medical textbooks) to enhance accuracy. However, performance tradeoffs across different RAG configurations—many of which exponentially increase computational cost—remain poorly characterized. Methods: In total, 120 lightweight, open-source RAG configurations were evaluated across 40 plastic surgery-focused question-answering tasks (20 single-hop, 20 multi-hop), spanning multiple subspecialties (4800 total evaluations). Configurations varied by base LLM (Phi-3-mini-128k-instruct vs. BioMistral-7B), embedding model, database size, chunk size, and query hop type. Performance was assessed using semantic similarity (Ragas) to physician-validated reference answers. Performance was analyzed using linear mixed-effects regression with query as a random effect and fixed and interaction effects selected via likelihood testing and AIC. Results: High performance was achievable using lightweight, open-source models. While BioMistral-7B had high mean sematic similarity under specific configurations (mean semantic similarity up to 0.786), Phi-3-mini-128k-instruct demonstrated more consistent performance across query complexity. Larger database sizes significantly improved semantic similarity, with the largest gain at intermediate sizes (e.g., size 5: +0.043, p = 0.001). Embedding choice had a strong effect, with bge-large-en-v1.5 improving performance (p = 0.0016) and Bio_ClinicalBERT markedly reducing it (p < 0.001). Multi-hop queries substantially reduced performance (p < 0.001), though this effect was attenuated for Phi-3-mini-128k-instruct via a strong model × hop-type interaction (p < 0.001). Conclusions: RAG systems for plastic surgery do not require large proprietary models, as performance depends on configuration choices and interaction effects rather than isolated components. With advancements, predictive modeling may enable resource-efficient, safe deployment of clinical RAG systems.

Abstract Despite growing interest in AI for pancreatic cyst management, no prior study has systematically investigated how lesion characteristics influence model behavior or provided a comprehensive explainability analysis in IPMN malignancy risk stratification. We present the first multi-center study integrating explainable AI with uncertainty quantification to evaluate how cyst type, size, and location affect deep learning performance in IPMN malignancy prediction. Our retrospective study analyzed 170 IPMNs from seven centers using a radiomics-deep learning fusion model. Cases were stratified by dysplasia grade, IPMN type, size, and location; with model interpretability assessed using SHAP, LIME, Grad-CAM, and uncertainty quantification. Overall accuracy was 67.1%, with model uncertainty lower for correct versus incorrect predictions (0.72 vs 0.78, p < 0.001). On subgroup analysis: Low-Risk lesion accuracy exceeded High-Risk (81.3% vs. 42.9%, p < 0.001), BD accuracy was inversely correlated with cyst size while MD remained stable, and body and whole-pancreas cysts stratified more accurately than tail cysts (77.8%, 76.6% vs. 47.8%). AI models face significant, previously uncharacterized performance drops when evaluating complex high-risk IPMNs. Our study shows that integrating uncertainty quantification can successfully flag unreliable predictions, enabling a selective-prediction framework where clinicians can confidently rely on AI for low-risk, low-uncertainty lesions while deferring high-uncertainty cases to expert review.

Abstract Background The association between long-term use of benzodiazepines (BDZs) and cognitive function in patients with schizophrenia has not been fully characterized. This study aimed to investigate the relationship between long-term BDZs use and cognitive impairment in schizophrenia. Methods Data were derived from an observational study of treatment with atypical antipsychotics in Chinese patients with schizophrenia (SALT-C). Fifty-seven patients with long-term use of BDZs (≥ 60 days) were included, and 57 BDZs non-users were matched using propensity scores for age, sex, disease duration, and atypical antipsychotics received. Cognitive impairment was assessed using the Montreal Cognitive Assessment (MoCA); psychotic symptoms, illness severity, and psychosocial functioning were assessed with the Positive and Negative Syndrome Scale (PANSS), the Clinical Global Impression-Severity (CGI-S) scale, and the Personal and Social Performance (PSP) scale. Between-group comparisons were performed using independent-samples t-tests or Mann–Whitney U tests. Multivariate analysis was performed using binary logistic regression. Results The long-term BDZs user group showed a significantly lower total MoCA score [16.02 (6.69)] than the BDZs non-user group [19.33 (7.11)] (P = 0.012). No significant between-group differences were observed in total scores on the PANSS, CGI-S, or PSP (all P > 0.05). Long-term BDZs use was independently associated with cognitive impairment (OR = 7.728, P = 0.017), whereas years of education was negatively associated with cognitive impairment (OR = 0.740, P = 0.007). Conclusions Long-term BDZs use is associated with poorer cognitive performance in patients with schizophrenia, suggesting that cognitive function should be monitored during BDZs treatment in this population.

Abstract Rare Mendelian disorders affect 300-400 million people globally. Although genetic testing has become widely adopted, gene-specific evidence for tailored variant interpretation remains scattered across resources. We present Gene Portals, a framework for gene-centered multimodal knowledge bases that co-localize expert-harmonized clinical data, functional assays, population variation, structural annotations and gene-specific ACMG/AMP specifications within a single resource. A modular interface integrates this unified evidence with VCEP-refined ACMG specifications to enable automated gene-specific variant classification, infer molecular mechanisms, and support cross-gene analyses. We demonstrate the framework’s utility across five Gene portals spanning eleven neurodevelopmental disorder-associated genes, integrating data from 4,423 individuals with 2,838 unique variants, 36,149 ClinVar submissions, and 1,044 expert-curated molecular readouts. By organizing evidence that is otherwise dispersed across multiple sources into a unified, queryable framework, the SCN, GRIN, CACNA1A, SATB2 and SLC6A1 Gene Portals became widely used community resources and provide an extensible template for standardized rare-disease variant interpretation and mechanism-aware discovery.