Machine Learning Approaches to Cancer Diagnostics in Humans Based on Laser Raman Microspectrometry of Human Body Fluids

Abstract

Near-infrared Raman spectroscopy is a spectroscopic technique capable of providing fingerprint-type information on biochemical molecules. For the early detection of cancer, specific biomarkers, e.g., biofluids’ biomarkers, need to be detected with high sensitivity. This enhances diagnostic accuracy in detecting biochemical fingerprints that would point to onset of cancer development. The aim of this study was to test and evaluate novelized machine learning techniques for detection and identification of trace biomarker alterations in saliva and blood pointing to the onset and progression of leukemia and breast cancers via a laser Raman spectral analysis approach. The spectral measurements were done in 393-2063 cm-1 region, based on a 785 nm excitation laser. The spectral data analysis were done in the 500-1800 cm-1 region; the considered fingerprint region for biological specimens. Trace biomarkers were studied by analysis of intermediate and higher-order principal components. The utility of intermediate and higher-order principal components in revealing trace biochemical alterations (trace biomarkers) in biological samples was first experimented on prostatic cells’ spectra data. The statistical relevance of principal components were determined by the use of the two-sample t-test and the effect size statistical criteria. For breast cancer and leukemia studies, the concentrations of trace biomarkers were estimated using the partial least squares regression model applied to the spectra of pure compounds and the biofluids spectrum. Whole blood and saliva simulates spiked with prepared concentrations of the various biochemical components ranging from 1 ppm to 500 ppm were used for for method development. Then, various optimized machine learning techniques that included independent component analysis (ICA), multidimensional scaling (MDS), partial least square discriminant analysis (PLS-DA), kernel density estimators, support vector machines (SVM), and backpropagation neural networks (BPNN) were utilized to analyze and classify the blood and saliva trace biomarkers’ Raman spectra from healthy and diseased subjects. Results using pairwise comparison of mean intensity (peak intensity ratios) and multivariate statistical techniques disclosed that biochemical changes of proteins, lipids, and nucleic acid components can be associated with prostate cancer, breast cancer, and leukemia progression. Four prominent regions: cytosine / guanine (566 ± 0.70 cm-1), glycerol (630 cm-1), saccharides (1370 ± 0.86 cm-1), tryptophan (1618 ± 1.73 cm-1); and six subtle regions: phospholipids (1076 cm-1), amide III (1232, 1234 cm-1), amide III (1276, 1278 cm-1),

phospholipids / nucleic acids (1330, 1333 cm-1), lipids (1434, 1442 cm-1), amide II (1471, 1479 cm-1) were identified, which can be regarded as useful biomarkers for prostate cancer diagnosis. Six spectral bands were determined: glycerol (589 cm-1), tryptophan / phosphatidylinositol (594 cm-1), glutamate / tryptophan (630 cm-1), glutamate (1626 cm-1), glycine / valine (1630 cm-1), and amide I / β-carotene (1638 cm-1) which can be regarded as new biomarkers of breast cancer in the blood-based breast cancer spectroscopy. The fitting model revealed that trace proteins, nucleic acids, and lipid biochemicals in blood and saliva increased with breast malignancy, whereas amounts of glycogen decreased with progression of breast malignancy. For blood samples, the determined concentrations of proteins, saccharides, amino acids, nucleic acids and lipids components in diseased patients were in the range of 237.82-384.96 ppm, 36.4-84.3 ppm, 14.31-83.69 ppm, 66.4-96.8 ppm, and 71.95-297 ppm, respectively, whereas respective concentrations in control samples were 233.86 ppm, 73.7 ppm, 10.48 ppm, 62.1 ppm, and 18-190 ppm. For saliva samples, concentrations of 62.5-126.3 ppm, 11.5-33.9 ppm, 4.90-20.6 ppm, 7.60-9.16 ppm, and 359.6 ppm representing trace proteins, saccharides, amino acids, nucleic acids and lipids in diseased patients were obtained. The respective concentrations in control samples were 27.7 ppm, 33.9 ppm, 2.17-3.66 ppm, 7.35 ppm, and 43.9-145.2 ppm. The quantitative analysis based on the selected trace biomarker regions suggested that biochemical changes of proteins and membranous lipids increased with leukemia malignancy whereas biochemical changes of nucleic acids, glycogen, and non-membranous lipids decreased with leukemia malignancy. For blood samples, the determined concentrations of proteins, saccharides, amino acids, nucleic acids and lipids components in diseased patients were 6.14 ppm, 2.8 ppm, 1.89-11.1 ppm, 32.25 ppm, and 2.21-3.9135 ppm, respectively, whereas respective concentrations in control samples were 4.04 ppm, 2.72 ppm, 2.29-14.7 ppm, 15.61 ppm, and 4.32- 7.1565 ppm. For saliva samples, concentrations of 8.737 ppm, 7.82 ppm, 15.88-17.80 ppm, 5.077 ppm, 0.282-3.645 ppm representing trace proteins, saccharides, amino acids, nucleic acids and lipids in diseased patients were obtained. The respective concentrations in control samples were 11.39 ppm, 14.90 ppm, 1.72-5.04 ppm, 1.069 ppm, and 1.81-4.769 ppm. The cross-validated models utilized to analyze and classify the blood and saliva Raman spectra from healthy subjects, breast tumor patients, and leukemia patients yielded diagnostic sensitivities of 46% to 100%, as well as specificities of 71% to 100%. Although the number of samples involved in this study were few, the results demonstrate that analysis of Raman spectra of blood and saliva using optimized machine learning diagnostic algorithms has great potential for the noninvasive and label-free detection of breast cancer and leukemia.