• HOME
  • NEWS
  • EXPLORE
    • CAREER
      • Companies
      • Jobs
    • EVENTS
    • iGEM
      • News
      • Team
    • PHOTOS
    • VIDEO
    • WIKI
  • BLOG
  • COMMUNITY
    • FACEBOOK
    • INSTAGRAM
    • TWITTER
Sunday, November 22, 2026
BIOENGINEER.ORG
No Result
View All Result
  • Login
  • HOME
  • NEWS
  • EXPLORE
    • CAREER
      • Companies
      • Jobs
        • Lecturer
        • PhD Studentship
        • Postdoc
        • Research Assistant
    • EVENTS
    • iGEM
      • News
      • Team
    • PHOTOS
    • VIDEO
    • WIKI
  • BLOG
  • COMMUNITY
    • FACEBOOK
    • INSTAGRAM
    • TWITTER
  • HOME
  • NEWS
  • EXPLORE
    • CAREER
      • Companies
      • Jobs
        • Lecturer
        • PhD Studentship
        • Postdoc
        • Research Assistant
    • EVENTS
    • iGEM
      • News
      • Team
    • PHOTOS
    • VIDEO
    • WIKI
  • BLOG
  • COMMUNITY
    • FACEBOOK
    • INSTAGRAM
    • TWITTER
No Result
View All Result
Bioengineer.org
No Result
View All Result
Home NEWS Science News Health

Machine Learning Model Predicts Pathological Response to Neoadjuvant Lung Cancer Immunotherapy

by
November 21, 2026
in Health
Reading Time: 6 mins read
0
Machine Learning Model Predicts Pathological Response to Neoadjuvant Lung Cancer Immunotherapy

Machine Learning Model Predicts Pathological Response to Neoadjuvant Lung Cancer Immunotherapy

Share on FacebookShare on TwitterShare on LinkedinShare on RedditShare on Telegram

A predictive model integrating programmed death-ligand 1 expression and tumor mutational burden accurately estimated pathological complete response in patients undergoing neoadjuvant immunotherapy for resectable lung cancer, according to research published in the International Journal of Clinical Pharmacy. Neoadjuvant immunotherapy has shown promising antitumor efficacy in resectable lung cancer, yet clinical responses remain highly variable across individual patients. While immune checkpoint inhibitors directed against the programmed death-1 and programmed death-ligand 1 pathways are increasingly combined with chemotherapy prior to surgical resection, a substantial proportion of patients experience limited therapeutic benefit while remaining exposed to immune-related toxicities, disease progression, or postponed surgery. Programmed death-ligand 1 expression and tumor mutational burden represent two established molecular biomarkers of immunotherapy responsiveness, but models combining both markers alongside baseline clinical characteristics have remained insufficiently evaluated. To address this predictive challenge, investigators developed and compared a traditional multivariable logistic regression nomogram and a machine learning random forest model to quantify the probability of achieving a complete pathological response prior to surgery.

The retrospective investigation evaluated 232 consecutive patients with resectable lung cancer treated with neoadjuvant immunotherapy followed by surgical resection between January 2023 and December 2024 at Henan Provincial Chest Hospital. This primary cohort served as the model development population and was randomly divided into a training cohort of 162 patients and an internal validation cohort of 70 patients using a seven-to-three random allocation scheme. To evaluate the broader transportability and temporal generalizability of the resulting predictive architectures, the investigators enrolled an independent temporal validation cohort comprising 115 consecutive patients who met identical eligibility criteria and underwent treatment at the same institution between January and December 2025. Eligible individuals had pathologically confirmed, clinically resectable stage II to IIIB lung cancer staged according to the eighth edition of the TNM staging system, had received preoperative immunotherapy, possessed resected surgical specimens with complete clinical and pathological documentation, and had tissue available for molecular profiling. Patients with concurrent malignancies, prior exposure to immunotherapy or targeted agents, autoimmune conditions, or organ dysfunction were excluded.

Preoperative assessment encompassed demographic characteristics, lifestyle factors, comorbid conditions, and laboratory measurements recorded in electronic medical archives. Evaluated clinical features included age, sex, smoking history defined as at least 100 cigarettes per year, histological classification including adenocarcinoma and squamous cell carcinoma, clinical disease stage, Eastern Cooperative Oncology Group performance status, tumor anatomical location, tumor size, and regional lymph node metastasis status. Laboratory evaluations included serum carcinoembryonic antigen quantification measured through chemiluminescent immunoassay with an upper reference threshold of 5.0 nanograms per milliliter. Molecular evaluation of pre-treatment tumor tissue evaluated programmed death-ligand 1 expression by immunohistochemistry utilizing the SP263 assay kit, with staining evaluated independently by two consensus pathologists and categorized by tumor proportion score as negative below 1 percent, low between 1 and 49 percent, or high at 50 percent or greater. Tumor mutational burden was quantified via next-generation sequencing across a targeted 422-gene panel on an Illumina NovaSeq 6000 platform, defining high mutational burden as at least 10 non-synonymous somatic mutations per megabase.

Neoadjuvant therapeutic regimens across the studied cohorts incorporated either immune checkpoint inhibitor monotherapy or checkpoint inhibitors paired with platinum-based doublet chemotherapy selected through multidisciplinary team evaluation. Administered monotherapy agents included intravenous atezolizumab at 1200 milligrams every three weeks, durvalumab at 1500 milligrams every four weeks, nivolumab at 240 milligrams every two weeks or 360 milligrams every three weeks, or pembrolizumab at 200 milligrams every three weeks for two to four cycles. Combination therapy coupled checkpoint inhibition with histology-specific platinum regimens, using pemetrexed with cisplatin or carboplatin for patients with adenocarcinoma and paclitaxel with platinum agents for patients diagnosed with squamous cell carcinoma. Following two to four neoadjuvant treatment cycles, surgical resections were executed, and the primary pathological endpoint was evaluated approximately four weeks after completing neoadjuvant therapy. Pathological complete response was formally defined under International Association for the Study of Lung Cancer standards as the complete absence of viable malignant cells in both the primary resected tumor and sampled regional lymph nodes.

Statistical modeling procedures utilized SPSS Statistics version 26.0 and R version 4.2.0, applying univariate logistic regression analyses to screen potential clinical, pathological, and molecular predictors of pathological complete response. Variables demonstrating significant association at an alpha threshold of 0.05 were subsequently entered into multivariable logistic regression models to establish independent predictors. The multivariable analysis demonstrated that clinical stage II disease yielded an odds ratio of 0.54 with a 95 percent confidence interval spanning 0.31 to 0.94, while the absence of lymph node metastasis yielded an odds ratio of 0.35 with a 95 percent confidence interval from 0.16 to 0.77. Concurrent neoadjuvant chemoimmunotherapy produced a substantial independent association with response, exhibiting an odds ratio of 4.00 and a 95 percent confidence interval between 1.54 and 10.39. Higher programmed death-ligand 1 expression independently predicted response with an odds ratio of 1.62 and a confidence interval of 1.03 to 2.54, while higher tumor mutational burden yielded an odds ratio of 3.07 with a confidence interval of 1.42 to 6.65.

Using the identified independent predictors, the researchers constructed a graphical nomogram through the rms package in R and assembled an ensemble random forest model via the randomForest package. The random forest architecture was constructed with 500 decision trees, five variables randomly sampled at each prospective split, and bootstrap resampling with replacement, incorporating five-fold cross-validation during training to restrict overfitting. Variable importance assessments based on the mean decrease in Gini impurity established that programmed death-ligand 1 expression and tumor mutational burden represented the two most influential predictors within the random forest algorithm. These molecular biomarkers were followed in predictive importance by the specific neoadjuvant treatment regimen, regional lymph node metastasis status, and clinical disease stage. In the model development cohort, median patient age was 63 years, with 160 men and 72 women, while the temporal validation cohort had a median age of 61 years with 68 men and 47 women, demonstrating broadly similar clinical characteristics aside from a higher proportion of nodal metastases in the temporal cohort.

Comparative discriminative evaluations demonstrated consistent performance advantages for the machine learning architecture over standard regression modeling. In the training set of 162 patients, the random forest model achieved an area under the receiver operating characteristic curve of 0.866, significantly outperforming the nomogram, which produced an area under the curve of 0.782, with a statistically significant difference at p less than 0.001. When evaluated within the internal validation cohort of 70 patients, the random forest retained superior discriminative capacity, yielding an area under the curve of 0.769 compared with 0.669 for the nomogram, also reaching statistical significance at p less than 0.001. Beyond discriminative area metrics, the random forest model delivered superior overall classification parameters, registering higher diagnostic accuracy, specificity, and precision across both internal evaluation sets. Although the nomogram exhibited greater sensitivity for identifying potential responders, the random forest model achieved a lower rate of false-positive classifications, indicating that each algorithmic architecture exhibited distinct predictive characteristics for clinical response estimation.

Model calibration was rigorously appraised across cohorts using dedicated calibration curves to measure concordance between predicted probabilities and observed pathological outcomes. Both the random forest model and the regression nomogram maintained good calibration in the initial training cohort. In the internal validation cohort, the nomogram demonstrated a modest reduction in calibration performance, recording a mean absolute error of 0.067, whereas the random forest model maintained acceptable calibration with minor divergence observed only at higher predicted probability levels. When tested in the temporally independent validation cohort of 115 consecutive patients treated in 2025, the random forest model demonstrated acceptable discrimination, yielding an area under the receiver operating characteristic curve of 0.757 with a 95 percent confidence interval spanning 0.682 to 0.902, accompanied by a sensitivity of 82.6 percent and a specificity of 76.2 percent. However, calibration curves in this temporal external cohort showed noticeable calibration drift, producing a Brier score of 0.139, a calibration intercept of negative 0.467, and a calibration slope of 0.183.

Biological mechanisms detailed by the authors support the complementary utility of combining programmed death-ligand 1 expression with genomic tumor mutational burden. Programmed death-ligand 1 expression reflects an active adaptive mechanism of tumor immune evasion through engagement of inhibitory programmed death-1 receptors on cytotoxic T cells. In contrast, tumor mutational burden functions as an indirect genomic indicator of overall tumor immunogenicity, reflecting the frequency of somatic non-synonymous mutations capable of generating novel neoantigens that stimulate host immune recognition. Integrating these orthogonal molecular characteristics provides a more comprehensive representation of the local tumor microenvironment than either biomarker evaluated in isolation. Furthermore, cytotoxic chemotherapy reduces tumor mass while promoting immunogenic cell death and tumor antigen release, providing a biological rationale for why combining chemotherapy with immune checkpoint inhibitors produced a fourfold increase in the likelihood of achieving complete pathological regression compared to checkpoint inhibitor monotherapy among resectable cases.

Despite promising discriminative performance, the study authors emphasized multiple methodological limitations that necessitate cautious interpretation. The retrospective, single-center design conducted at a single hospital introduces potential selection bias and unmeasured confounding variables. The suboptimal calibration observed in the temporal validation cohort highlights the influence of inter-cohort heterogeneity, including variations in chemotherapy backbones, specific checkpoint inhibitors, and evolving treatment intervals across distinct treatment periods. The study relied on a single immunohistochemical assay platform for programmed death-ligand 1 and a specific 422-gene sequencing panel, meaning threshold definitions may vary across clinical laboratories. Furthermore, the analysis excluded additional potentially relevant biological indicators such as tumor-infiltrating lymphocytes, microsatellite instability, and oncogenic driver mutations, while focusing entirely on pathological complete response without assessing long-term survival outcomes including disease-free survival and overall survival. Consequently, the investigators concluded that this random forest model must be regarded as exploratory until large-scale prospective multicenter validation and algorithmic recalibration are completed.

Subject of Research: Predictive modeling of pathological complete response after neoadjuvant immunotherapy in resectable lung cancer

Article Title: Development and validation of a PD-L1 and tumor mutational burden-based predictive model for pathological complete response after neoadjuvant immunotherapy in resectable lung cancer

Article References: Wu, X., Jiang, D., & Li, B. (2026). Development and validation of a PD-L1 and tumor mutational burden-based predictive model for pathological complete response after neoadjuvant immunotherapy in resectable lung cancer. International Journal of Clinical Pharmacy, 48(5), 2082-2093. https://doi.org/10.1007/s11096-026-02191-3

Image Credits: AI Generated

DOI: 10.1007/s11096-026-02191-3

Keywords: neoadjuvant immunotherapy, resectable lung cancer, pathological complete response, PD-L1 expression, tumor mutational burden, random forest model, nomogram, non-small cell lung cancer, chemoimmunotherapy, biomarker integration, temporal validation, receiver operating characteristic, Henan Provincial Chest Hospital, SP263 immunohistochemistry, next-generation sequencing, clinical stage II, lymph node metastasis

News Source: Nathaniel Bowman. (November 21, 2026). Machine Learning Model Predicts Pathological Response to Neoadjuvant Lung Cancer Immunotherapy. Scienmag.

Tags: Biomarker integrationchemoimmunotherapyclinical stage IIHenan Provincial Chest Hospitallymph node metastasisNeoadjuvant immunotherapynext-generation sequencingnomogramnon-small cell lung cancerPathological Complete ResponsePD-L1 expressionrandom forest modelreceiver operating characteristicresectable lung cancerSP263 immunohistochemistrytemporal validationtumor mutational burden
Share12Tweet7Share2ShareShareShare1

Related Posts

Pyrogallic Acid Restores Doxycycline Susceptibility in Multidrug-Resistant Enterotoxigenic Escherichia Coli

Pyrogallic Acid Restores Doxycycline Susceptibility in Multidrug-Resistant Enterotoxigenic Escherichia Coli

November 22, 2026
Adjunctive Phage Therapy Achieves Eighty-Two Percent Remission in Refractory Musculoskeletal Infections

Adjunctive Phage Therapy Achieves Eighty-Two Percent Remission in Refractory Musculoskeletal Infections

November 22, 2026

Single Hyaluronic Acid Injection Matches Three-Dose Therapy for Knee Osteoarthritis

November 22, 2026

Fanconi Anemia Pathway Suppression of NHEJ Prevents MLL-Rearranged Leukemia Progression

November 22, 2026

POPULAR NEWS

  • Review Maps Multistep Host Networks Directing Human Papillomavirus Entry and Trafficking

    Review Maps Multistep Host Networks Directing Human Papillomavirus Entry and Trafficking

    29 shares
    Share 12 Tweet 7
  • Adjunctive Phage Therapy Achieves Eighty-Two Percent Remission in Refractory Musculoskeletal Infections

    29 shares
    Share 12 Tweet 7
  • Single Hyaluronic Acid Injection Matches Three-Dose Therapy for Knee Osteoarthritis

    29 shares
    Share 12 Tweet 7
  • Fanconi Anemia Pathway Suppression of NHEJ Prevents MLL-Rearranged Leukemia Progression

    29 shares
    Share 12 Tweet 7

About

We bring you the latest biotechnology news from best research centers and universities around the world. Check our website.

Follow us

Recent News

Pyrogallic Acid Restores Doxycycline Susceptibility in Multidrug-Resistant Enterotoxigenic Escherichia Coli

Review Maps Multistep Host Networks Directing Human Papillomavirus Entry and Trafficking

Adjunctive Phage Therapy Achieves Eighty-Two Percent Remission in Refractory Musculoskeletal Infections

Subscribe to Blog via Email

Enter your email address to subscribe to this blog and receive notifications of new posts by email.

Join 89 other subscribers
  • Contact Us

Bioengineer.org © Copyright 2023 All Rights Reserved.

Welcome Back!

Login to your account below

Forgotten Password?

Retrieve your password

Please enter your username or email address to reset your password.

Log In
No Result
View All Result
  • Homepages
    • Home Page 1
    • Home Page 2
  • News
  • National
  • Business
  • Health
  • Lifestyle
  • Science

Bioengineer.org © Copyright 2023 All Rights Reserved.