MSMelanie Spears

Melanie Spears

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Melanie Spears has a total of 84 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Spatially resolved gene expression and tumour-microenvironment interactions across malignant transformation in IDH-mutant glioma

    by Alyona Ivanova, Shamini Ayyadhury, David G Munoz, Megan Hopkins, Farzaneh Aboualizadeh, Melanie Spears, Megan Wu, Trevor Pugh, Roel G W Verhaak, Daniel P Cahill, et al.

    Isocitrate dehydrogenase (IDH)-mutant gliomas represent most lower-grade diffuse gliomas in young adults. Although IDH-mutant gliomas initially behave indolently relative to IDH-wildtype counterparts, they undergo inevitable malignant transformation. The mechanisms driving this progression remain poorly defined. We analyzed a unique longitudinal cohort of two patients with IDH-mutant diffuse gliomas, each with matched WHO grade 2, 3, and 4 tumours. Spatial transcriptomics (NanoString GeoMx and 10X Visium), combined with differential expression, spatially variable gene analysis, and gene ontology enrichment, were used to define transcriptional programs underlying progression. Unexpectedly, across both patients, higher-grade tumours demonstrated reduced proliferation at the infiltrative margin as measured by Ki67. Spatial transcriptomic analysis identified an enrichment of invasion-associated genes associated with increasing tumour grade. In an independent bulk RNA-seq cohort, expression of this spatially derived program did not outperform grade-only PFS or OS analysis. These observations underscore the need to develop new metrics that account for spatial evolution that occurs across progression, but escapes detection in bulk analyses and provide the rationale for prioritizing migration-, cytoskeletal- and stress-associated programs for further investigation.

  • Open Access

    Transfer Learning and Machine Learning for Training Five-Year Survival Prognostic Models in Early Breast Cancer: Development and Validation Study

    by Lisa Pilgram, Kai Yang, Ana-Alicia Beltran-Bless, Gregory R Pond, Lisa Vandermeer, John Hilton, Marie-France Savard, Andreanne LeBlanc, Lois Shepherd, Bingshu Chen, et al.

    Background: Prognostic information is essential for decision-making in breast cancer management. In recent years, trials and clinical practice have emphasized genomic prognostication tools, despite clinicopathological methods being more affordable and accessible. PREDICT v3 is one such tool with promising results across cohorts. Advances in machine learning (ML), transfer learning, and ensemble methods provide opportunities to enhance these approaches, especially where missing data and model assumptions differ across diverse populations. Objective: This study evaluates the potential to improve survival prognostication in breast cancer. More precisely, we compare de novo ML, transfer learning from the pretrained prognostication model PREDICT v3, and a stacked ensemble approach. Methods: Data from the MA.27 trial (NCT00066573) were used for model training, with external validation on data from the Tamoxifen Exemestane Adjuvant Multinational trial (NCT00279448 and NCT00032136) and a US Surveillance, Epidemiology, and End Results cohort. Transfer learning was applied by re-estimating the parameters of the pretrained prognostic tool PREDICT v3. De novo ML included random survival forests and extreme gradient boosting, and the ensemble was implemented using weighted linear stacking of model predictions. Internal and external validation was assessed in terms of the integrated calibration index and discrimination. Shapley Additive Explanations values were used to explain model predictions and decision-curve analysis to facilitate the interpretation of performance differences. Results: Transfer learning, de novo random survival forest, and the stacked ensemble improved calibration in MA.27 over the pretrained model (integrated calibration index reduced from 0.042 in PREDICT v3 to ≤0.007) while discrimination remained comparable (AUROC increased from 0.738 in PREDICT v3 to 0.744-0.799). In decision-curve analysis, these approaches demonstrated consistently positive net benefit across clinically relevant thresholds, while PREDICT v3 lost net benefit beyond 7.5% predicted risk. Invalid PREDICT v3 predictions were observed in 23.8% to 25.8% of MA.27 individuals due to missing information. In contrast, ML models and the stacked ensemble predicted survival despite missing data. Across all models, patient age, nodal status, pathological grading, and tumor size had the highest Shapley Additive Explanations values, indicating their importance for survival prognostication. External validation in the US Surveillance, Epidemiology, and End Results cohort confirmed the benefits of transfer learning, RSF, and ensemble in terms of calibration while maintaining discrimination at comparable levels. In contrast, generalizability was limited in the Tamoxifen Exemestane Adjuvant Multinational trial, a cohort with a substantially different distribution of clinicopathological characteristics. Conclusions: This study demonstrates that transfer learning, de novo RSF, and a stacked ensemble can improve prognostication compared with the pretrained PREDICT v3, particularly in the presence of missing or uncertain inputs. Transportability may be limited in cohorts with different clinicopathological profiles, requiring local validation before clinical deployment. Ultimately, better survival estimation can provide more meaningful guidance in breast cancer care. Trial Registration: ClinicalTrials.gov NCT00066573; https://clinicaltrials.gov/study/NCT00066573, NCT00279448; https://clinicaltrials.gov/study/NCT00279448, NCT00032136; https://clinicaltrials.gov/study/NCT00032136


  • Open Access

    Transfer Learning and Machine Learning for Training Five-Year Survival Prognostic Models in Early Breast Cancer: Development and Validation Study (Preprint)

    by Lisa Pilgram, Kai Yang, Ana-Alicia Beltran-Bless, Gregory R Pond, Lisa Vandermeer, John Hilton, Marie-France Savard, Andreanne LeBlanc, Lois Shepherd, Bingshu Chen, et al.

    BACKGROUND Prognostic information is essential for decision-making in breast cancer management. In recent years, trials and clinical practice have emphasized genomic prognostication tools, despite clinicopathological methods being more affordable and accessible. PREDICT v3 is one such tool with promising results across cohorts. Advances in machine learning (ML), transfer learning, and ensemble methods provide opportunities to enhance these approaches, especially where missing data and model assumptions differ across diverse populations. OBJECTIVE This study evaluates the potential to improve survival prognostication in breast cancer. More precisely, we compare de novo ML, transfer learning from the pretrained prognostication model PREDICT v3, and a stacked ensemble approach. METHODS Data from the MA.27 trial (NCT00066573) were used for model training, with external validation on data from the Tamoxifen Exemestane Adjuvant Multinational trial (NCT00279448 and NCT00032136) and a US Surveillance, Epidemiology, and End Results cohort. Transfer learning was applied by re-estimating the parameters of the pretrained prognostic tool PREDICT v3. De novo ML included random survival forests and extreme gradient boosting, and the ensemble was implemented using weighted linear stacking of model predictions. Internal and external validation was assessed in terms of the integrated calibration index and discrimination. Shapley Additive Explanations values were used to explain model predictions and decision-curve analysis to facilitate the interpretation of performance differences. RESULTS Transfer learning, de novo random survival forest, and the stacked ensemble improved calibration in MA.27 over the pretrained model (integrated calibration index reduced from 0.042 in PREDICT v3 to ≤0.007) while discrimination remained comparable (AUROC increased from 0.738 in PREDICT v3 to 0.744-0.799). In decision-curve analysis, these approaches demonstrated consistently positive net benefit across clinically relevant thresholds, while PREDICT v3 lost net benefit beyond 7.5% predicted risk. Invalid PREDICT v3 predictions were observed in 23.8% to 25.8% of MA.27 individuals due to missing information. In contrast, ML models and the stacked ensemble predicted survival despite missing data. Across all models, patient age, nodal status, pathological grading, and tumor size had the highest Shapley Additive Explanations values, indicating their importance for survival prognostication. External validation in the US Surveillance, Epidemiology, and End Results cohort confirmed the benefits of transfer learning, RSF, and ensemble in terms of calibration while maintaining discrimination at comparable levels. In contrast, generalizability was limited in the Tamoxifen Exemestane Adjuvant Multinational trial, a cohort with a substantially different distribution of clinicopathological characteristics. CONCLUSIONS This study demonstrates that transfer learning, de novo RSF, and a stacked ensemble can improve prognostication compared with the pretrained PREDICT v3, particularly in the presence of missing or uncertain inputs. Transportability may be limited in cohorts with different clinicopathological profiles, requiring local validation before clinical deployment. Ultimately, better survival estimation can provide more meaningful guidance in breast cancer care. CLINICALTRIAL ClinicalTrials.gov NCT00066573; https://clinicaltrials.gov/study/NCT00066573, NCT00279448; https://clinicaltrials.gov/study/NCT00279448, NCT00032136; https://clinicaltrials.gov/study/NCT00032136

  • Open Access

    Figure 2 from Continuous versus Standard Palbociclib Treatment and Molecular Profiling of Solid Tissues and Liquid Biopsies in the CCTG MA.38 Trial in Advanced Breast Cancer

    by Anil Abraham Joy, Nicholas Cheng, Karen A. Gelmon, Mihaela Mates, Christine Desbiens, Mark Clemons, Sara Taylor, Julie Lemieux, Angela DeLuca, Louis Gasparini, et al.

    Pretreatment diagnostic tumor somatic variants and transcriptional signatures associated with PFS. A, Frequency of variants across breast cancer driver genes in pretreatment tumor biopsies. Colors indicate variant classification (B) VAF of detected single-nucleotide variants across samples. C, HRs of PFS associated with the presence of somatic variants in driver genes. D–F, Kaplan–Meier curves of genetic variants associated with shorter PFS time in (D) TP53 and (E) CDH1. F–I, Kaplan–Meier curves of transcriptional signatures stratified by TP53 variant status associated with shorter PFS time for (F) HRD, (G) hypoxia, (H) androgen receptor, and (I) claudin low.

  • Open Access

    Supplementary Figure S9 from Continuous versus Standard Palbociclib Treatment and Molecular Profiling of Solid Tissues and Liquid Biopsies in the CCTG MA.38 Trial in Advanced Breast Cancer

    by Anil Abraham Joy, Nicholas Cheng, Karen A. Gelmon, Mihaela Mates, Christine Desbiens, Mark Clemons, Sara Taylor, Julie Lemieux, Angela DeLuca, Louis Gasparini, et al.

    Figure S9. Association between tumour-fraction, short to long fragment length ratios and progression-free survival time. A) Hazard ratio between individuals above and below each tumour fraction cut-off at baseline (green), W12 (red) and W24 (yellow). Filled in points indicate significant hazard ratios. B) Hazard ratios of tumour fraction change between baseline and W12 cfDNA associated with progression-free survival at various cut-offs. Filled in points indicate significant hazard ratios. C) Cell-free DNA fragment length distribution between individuals with shorter and longer progression free survival time of three months. D) Hazard ratio of short to long fragment length ratio on progression free survival time using various cut-offs to divide short and long. Colours indicate cfDNA sampling time at baseline (green), W12 (red) and at W24 (yellow). Filled in points are significant log-rank p-values fragment length cut-offs.

  • Open Access

    Supplementary Table 1 from Continuous versus Standard Palbociclib Treatment and Molecular Profiling of Solid Tissues and Liquid Biopsies in the CCTG MA.38 Trial in Advanced Breast Cancer

    by Anil Abraham Joy, Nicholas Cheng, Karen A. Gelmon, Mihaela Mates, Christine Desbiens, Mark Clemons, Sara Taylor, Julie Lemieux, Angela DeLuca, Louis Gasparini, et al.

    Supplementary Table 1. Breast Cancer 360 Gene Signatures and Description

  • Open Access

    Figure 3 from Continuous versus Standard Palbociclib Treatment and Molecular Profiling of Solid Tissues and Liquid Biopsies in the CCTG MA.38 Trial in Advanced Breast Cancer

    by Anil Abraham Joy, Nicholas Cheng, Karen A. Gelmon, Mihaela Mates, Christine Desbiens, Mark Clemons, Sara Taylor, Julie Lemieux, Angela DeLuca, Louis Gasparini, et al.

    Somatic variants detected in cfDNA associated with PFS time at baseline and first follow-up at week 12. A and B, Summary of genetic variants captured at (A) baseline prior to palbociclib treatment and at (B) week 12. Blue squares indicate the detection of a somatic variant in cfDNA for individuals. Barplot colors indicate variant classification. C and D, HR associated with PFS time for genetic variants detected at (C) baseline and at (D) week 12. Red points indicate gene variants significantly associated with PFS. E–I, Kaplan–Meier curves of mutations significantly associated with poor PFS time at (E–G) baseline in (E) SETD2, (F) DDR2, and (G) FGFR4 and (H and I) at week 12 in (H) TP53 and (I) ESR1. J and K, Kaplan–Meier curves of mutations associated with poor PFS in groups of genes in the (J) DNA repair pathway and (K) MAPK cascade.

  • Open Access

    Supplementary Figure S7 from Continuous versus Standard Palbociclib Treatment and Molecular Profiling of Solid Tissues and Liquid Biopsies in the CCTG MA.38 Trial in Advanced Breast Cancer

    by Anil Abraham Joy, Nicholas Cheng, Karen A. Gelmon, Mihaela Mates, Christine Desbiens, Mark Clemons, Sara Taylor, Julie Lemieux, Angela DeLuca, Louis Gasparini, et al.

    Figure S7. Summary of top mutated genes in solid tissue and longitudinal liquid biopsy samples.

  • Open Access

    Table 2 from Continuous versus Standard Palbociclib Treatment and Molecular Profiling of Solid Tissues and Liquid Biopsies in the CCTG MA.38 Trial in Advanced Breast Cancer

    by Anil Abraham Joy, Nicholas Cheng, Karen A. Gelmon, Mihaela Mates, Christine Desbiens, Mark Clemons, Sara Taylor, Julie Lemieux, Angela DeLuca, Louis Gasparini, et al.

    Table 2 from Continuous versus Standard Palbociclib Treatment and Molecular Profiling of Solid Tissues and Liquid Biopsies in the CCTG MA.38 Trial in Advanced Breast Cancer

  • Open Access

    Supplementary Figure S6 from Continuous versus Standard Palbociclib Treatment and Molecular Profiling of Solid Tissues and Liquid Biopsies in the CCTG MA.38 Trial in Advanced Breast Cancer

    by Anil Abraham Joy, Nicholas Cheng, Karen A. Gelmon, Mihaela Mates, Christine Desbiens, Mark Clemons, Sara Taylor, Julie Lemieux, Angela DeLuca, Louis Gasparini, et al.

    Supplementary Figure S6. Prognostic transcriptional signatures stratified by TP53 mutation and BC360 HRD expression status in solid tissues.