Mass Spectrometry and Machine Learning for the Diagnosis of B-Cell Lymphoma in Dogs
Principal Investigator: Priscila Serpa
DESCRIPTION (provided by applicant):
Lymphoma is the most common hematopoietic malignancy in dogs, accounting for up to approximately 25% of all canine neoplasms, with nearly two thirds being of B origin. In dogs, the distinction between B and T cell origin is relevant for the prognosis, since T-cell lymphomas are often more aggressive and less responsive to the treatment. Diagnostic procedures and tests for definitive diagnosis and phenotyping of B-cell lymphoma can be costly, time-consuming, and invasive. We propose using an instrument commonly found in microbiology diagnostic laboratories to provide a rapid and cheaper means to diagnose lymphoma in dogs (requiring only blood or a fine-needle tissue aspirate (FNA). Matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS) is an analytical technique routinely used to identify bacteria. My overall hypothesis is that MALDI-TOF MS in association with machine learning (ML) can accurately distinguish canine neoplastic from non-neoplastic B lymphocytes in blood and tissue aspirates and that the analysis through ML of protein spectra will increase the method's effectiveness. The broad objective of this project is to develop the use of MALDI-TOF MS to identify neoplastic B lymphocytes in two different biological matrices: fresh whole blood and smears from fine-needle tissue aspirates of lymphoid tissue. The study will be divided into two phases: (1) a MALDI development phase, to create the reference spectral libraries od neoplastic and non-neoplastic cells, and (2) an ML development phase, employing different ML techniques to improve diagnostic accuracy. If successful, a third phase will be necessary to validate the ML algorithm. In specific aim 1, we will develop spectral libraries using 4 sample types: i) lymphocytes isolated from venous blood of dogs with non-neoplastic, noninfectious conditions; ii) B-cell lymphoma cell line (CLBL-1); iii) smears of cytologically non-neoplastic lymphnodes; and iv) smears of lymph node aspirates with immunohistochemistry (IHC) or flow cytometry-confirmed B cell lymphoma. During this phase, cells should be lysed and proteins extracted for sample analysis. The intra-and inter-assay variability (precision) and analytical sensitivity and specificity will be determined. In aim 2, blood (prospective collection) and a larger cohort of archives smears of lymph nodes of dogs diagnosed with B-cell lymphoma and with non-neoplastic conditions will be used to expand these libraries and serve as a training and validation sets for ML. Today, dogs that are suspicious of having lymphoma usually pass through an FNA for a relatively cheap and quick cytologic diagnosis of lymphoma. However, cytology alone cannot provide a phenotype. Then, a second round of diagnostic procedures should be performed, usually biopsy for histopathology and IHC or a new FNA for flow cytometry and clonality tests, which all represent a longer time and a higher cost until the animal can start to be treated. Our idea on using MALDI-TOF MS is that the first FNA could be used for diagnosis and simultaneous phenotyping, representing a cheaper method and with shorter turnaround time. In future studies, we can evaluate the use of this method for detection of minimal residual disease and application for other types of cancer. Furthermore, since MALDI are available in several institutions, it could potentially become a widely available method of oncologic diagnosis.