Search

Background: Radiation therapy (RT) is frequently recommended for post-surgery treatment of early-stage breast cancer (BC) patients, though not all benefit. Clinical factors currently guide RT treatment decisions. This work presents a high-accuracy genomic machine learning (ML) model to predict RT-benefit in early-stage BC patients and a novel method for selecting genomic features for ML algorithms. Methods: Gene expression data from 477 early-stage BC patients treated with surgery and RT from the METABRIC cohort were obtained. Wilcoxon Rank Sum (Wilcox RS) test and Cox Proportional Hazards (Cox PH) were used to reduce the number of genes used to train 8 ML algorithms. Each ML algorithm was trained on a random subset of 80% of the data using 10-fold cross-validation and tested on the remaining 20% to assess its performance in predicting relapse status within 30 years. Results: The genomic data were reduced using Wilcox RS and Cox PH to a 1,596 gene set and a 977 gene set. These gene sets when used to train the 8 ML algorithms resulted in models that ranged in performance accuracies from 54.01% to 95.6%. The highest accuracies were obtained using Support Vector Machine (SVM977 - 93.41%, SVM1596 - 95.6%) and Neural Networks (NN977 - 92.31%, NN1596 - 93.41%). The accuracy of all models when tested on RT-untreated patients was 30-40% lower compared to RT-treated patients. SVM977 had the highest sensitivity of 91.09%. Members of the 977 gene set were enriched with genes involved in the cell cycle and differentiation as well as radiogenes. Conclusion: We developed an SVM model that used 977 differentially expressed genes as features that predicted RT-benefit in early-stage BC patients with 93.41% accuracy and 91.09% sensitivity. We also developed a novel genomic feature selection approach that used Wilcoxon RS followed by Cox PH that resulted in expression values of only 4% of all genes being used as features in the models. Citation Format: Kimberly Badal, Jerome Foster, Rajini Haraksingh, Melford John. Genomic machine learning model predicts radiation therapy benefit in early-stage breast cancer patients with high accuracy [abstract]. In: Proceedings of the 2021 San Antonio Breast Cancer Symposium; 2021 Dec 7-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2022;82(4 Suppl):Abstract nr P3-19-05.

BackgroundRadiation therapy (RT) is frequently recommended for post-surgery treatment of early-stage breast cancer (BC) patients, though not all benefit. Clinical factors currently guide RT treatment decisions. At present, models to predict RT-benefit predominantly use statistical methods with modest performance. In this paper we present a high-accuracy genomic Machine Learning (ML) model to predict RT-benefit in early-stage BC patients. We also present a novel method for selecting genomic features for training ML algorithms. MethodsGene expression data from 463 early-stage BC patients treated with surgery and RT from the METABRIC cohort were obtained. Wilcoxon Rank Sum (Wilcoxon RS) test and Cox Proportional Hazards (Cox PH) were used to reduce the number of genes used to train eight ML algorithms. ML algorithms were trained on 80% of data using 10-fold cross validation and tested on 20% of data to assess performance in predicting relapse status. Results Genome-wide gene expression data was reduced by 96% using Wilcoxon RS and Cox PH to a 1,596 gene set and a 977 gene set. These gene sets were used to train eight ML algorithms resulting in models that ranged in performance accuracies from 54.01% to 95.6%. Highest accuracies were obtained using Support Vector Machine (SVM977–93.41%, SVM1596–95.6%) and Neural Networks algorithms (NN977 – 92.31%, NN1596 – 93.41%). In RT-untreated patients, accuracies of all models were 30% to 40% lower compared to RT-treated patients. SVM977 had the highest sensitivity of 91.09%. Members of the 977 set were enriched with genes involved in cell cycle and differentiation as well as genes associated with radiosensitivity and radioresistance. Conclusion This study presents a novel genomic feature selection approach that used Wilcoxon RS followed by Cox PH to reduce the number of genes from genome-wide gene expression data used for training ML algorithms by 96%. This approach led to an SVM model that used the expression values of 977 genes to predict RT-benefit in early-stage BC patients with 93.41% accuracy. This work demonstrates that ML models can be clinically useful for predicting cancer patient outcomes.

Manifest and cluster files used in Genome Studio analysis of Illumina arrays. Lists the Illumina-supplied manifest and cluster files for each array that were used in Genome Studio analysis. These files were downloaded from http://support.illumina.com/array/downloads.html . (DOCX 57 kb)

Array detection of gold standard deletions > 100 kb in size. Summarizes the detection of the seven gold standard CNVs > 100 kb by the different arrays including the number of arrays that detected each large CNV, which arrays detected each CNV, and possible reasons why a particular large CNV may not be detectable by certain arrays. (DOCX 114 kb)

Cumulative frequencies of sizes of validated and non-validated CNV calls using platform specific algorithm. Cumulative frequencies of the sizes of validated CNVs are shown in red. Cumulative frequencies of sizes of non-validated CNVs are shown in green. CNV size is plotted on a log scale. Plots are shown for all arrays with more than 50 validated CNVs called using the platform specific algorithm. P-values were computed using a Mannâ Whitney U test that corrects for ties and uses a continuity correction. The p-values correspond to a one-sided hypothesis. (PDF 88 kb)

Summary: The variant call format (VCF) is a generic format for storing DNA polymorphism data such as SNPs, insertions, deletions and structural variants, together with rich annotations. VCF is usually stored in a compressed manner and can be indexed for fast data retrieval of variants from a range of positions on the reference genome. The format was developed for the 1000 Genomes Project, and has also been adopted by other projects such as UK10K, dbSNP and the NHLBI Exome Project. VCFtools is a software suite that implements various utilities for processing VCF files, including validation, merging, comparing and also provides a general Perl API. Availability: http://vcftools.sourceforge.net Contact: rd@sanger.ac.uk

J.B.S. Haldane proposed in 1947 that the male germline may be more mutagenic than the female germline. Diverse studies have supported Haldane's contention of a higher average mutation rate in the male germline in a variety of mammals, including humans. Here we present, to our knowledge, the first direct comparative analysis of male and female germline mutation rates from the complete genome sequences of two parent-offspring trios. Through extensive validation, we identified 49 and 35 germline de novo mutations (DNMs) in two trio offspring, as well as 1,586 non-germline DNMs arising either somatically or in the cell lines from which the DNA was derived. Most strikingly, in one family, we observed that 92% of germline DNMs were from the paternal germline, whereas, in contrast, in the other family, 64% of DNMs were from the maternal germline. These observations suggest considerable variation in mutation rates within and between families.

The 1000 Genomes Project was launched as one of the largest distributed data collection and analysis projects ever undertaken in biology. In addition to the primary scientific goals of creating both a deep catalogue of human genetic variation and extensive methods to accurately discover and characterize variation using new sequencing technologies, the project makes all of its data publicly available for community use. The project data coordination center has developed and deployed several tools to enable widespread data access.

Evolution, Gene Number, and Disease Slight variations in the numbers of copies of genes influence human disease and other characters. Variants can be hard to detect when they lie in heavily duplicated and widely similar regions of sequence known as “dark matter.” Sudmant et al. (p. 641 ) have methods to tease apart the duplicated regions to reveal singly unique nucleotide identifiers. These have turned out to be among the most variable seen in different human population groups—most notably among genes for neurodevelopment and neurological diseases. Such polymorphisms can be genotyped with specificity and may help us understand how variation in copy number may affect human evolution and disease.