Maldi-tof spectral fingerprints analysis by artificial neural networks of different types of pathological pain in experimental models
- Autores: Meritxell Deulofeu
- Directores de la Tesis: Pere Boadas Vaello (dir. tes.), Victòria Salvadó Martín (codir. tes.)
- Lectura: En la Universitat de Girona ( España ) en 2019
- Idioma: español
- Tribunal Calificador de la Tesis: Xavier Ribas Salamaña (presid.), Luz Romero Alonso (secret.), Héctor Gallart Ayala (voc.)
- Programa de doctorado: Programa de Doctorado en Biología Molecular, Biomedicina y Salud por la Universidad de Girona
- Materias:
- Texto completo no disponible (Saber más ...)
- Resumen
Pathological pain is widely prevalent in our society and it may be considered as a major health concern. There are different types of pathological pain which include both neuropathic and nociplastic pain. While examples of the first condition are spinal cord injury-induced central neuropathic pain and nerve chronic constriction injury-induced peripheral neuropathic pain, Fibromyalgia is one of the most common syndromes showing nociplastic symptoms. Nowadays, pathological pain is an important health concern for two different reasons: there are no effective treatments and current diagnosis protocols are only based on clinical criteria. Consequently, this unmet needs lead to the pathological pain chronification which in turn triggers the appearance of comorbid psychological diseases such as anxiety and depression, reducing dramatically the patients’ quality of life. Given this current situation, there is a need to find out new accurate diagnostic methods and also new pharmacological targets. Both neuropathic pain and fibromyalgia syndrome are characterised by exacerbated pain-responses like thermal hyperalgesia and mechanical allodynia, which further complicates the diagnosis of these diseases only based on patients’ symptoms. Since pathological pain is considered as a complex disease, the use of the classical laboratory tests such as immunohistochemical assays cannot be the best approach for pain studies. In contrast, the use of analytical techniques such as Matrix Assisted Laser Desorption Ionization Time of Flight (MALDI- TOF) mass spectrometry (MS) has several advantages such as its capability to simultaneously determine several molecules in a wide range of mass (from low molecular weight metabolites to peptides and proteins) and the possibility of characterising the fingerprint of a group of non-identified molecules. That is, we considered that using a profiling or fingerprint approach would be the better strategy to diagnose pathological pain since assuming that neuropathic and nociplastic pain molecular pattern are different, knowing the underlying identities of the different peaks is not necessary. However, following this strategy, data obtained with mass spectrometric techniques is characterized by having high complexity and requires of the application of statistical tools such as principal component analysis (PCA) in order to reduce their dimensionality facilitating both the interpretation and the evaluation of the data. The reduced mass spectrum data is then analysed by artificial neural networks (ANNs), consisting of a set of algorithms that are designed to recognize patterns, generating a classification output after being trained with a labelled dataset. Thus, the global objective of the present thesis was to develop a new methodology by combining MALDI-TOF MS and ANN analyses to obtain a model mainly based on the spectral fingerprints of biological samples of different pathological pain which can be used as a suitable diagnostic tool. To this end, central and peripheral neuropathic and fibromyalgia-like pains were induced in CD1 female mice, which developed both thermal hyperalgesia and mechanical allodynia. Serum from these animals were analysed with three different mass spectrometers, which differ in the mass analyser, and the mass spectrum data was then analysed using ANNs. Via comparison with the respective control groups, differences in mass spectra were observed in the low mass range (<1,000Da). A database with the most relevant signals found in the mass spectra which were selected using multivariate data tools was created for each model and each mass spectrometer data. ANN analyses of the different fingerprints allowed a good discrimination between the four pathological pain groups and their respective controls. Moreover, several differences were observed during the comparison between the mass spectra of different pathological pain regardless the mass spectrometer used. ANN analyses permitted a really good discrimination between serum samples of the different-nature pain animal model. Interestingly, the addition of other relevant variables such as functional pain outcomes or spinal cord glia reactivation improves the discriminatory/classification ability of the developed methodology. In addition, spinal cord tissue samples of all the different pain models were also analysed. Differences in the metabolomics pattern were also observed not only between pain groups when compared to controls but also between tissue samples of the different-nature pain animal models. The changes detected in the spinal cord tissue, although they may dependent on the condition inducing pain are not relevant for the diagnoses of pathological pain but are biomedically interesting because they could be new potential therapeutic targets under the different m/z signals. In summary, an innovative, simple and fast method for the detection and classification of pathological pain using serum mass spectra has been developed in the present thesis. The analysis of the serum using MALDI-TOF MS in combination with ANNs enables a methodology for the detection of pathological pain and its type through the analysis of their fingerprints without the need to identify the markers. Furthermore, the ability of this methodology to find new therapeutic targets has been suggested.
