Biomarkerji človeškega črevesnega mikrobioma za napovedovanje depresije

Primožič, Maša (author), Stres, Blaž (mentor)

Depresija je ena izmed vodilnih duševnih motenj na svetu in ima vrsto negativnih posledic za posameznika ter družbo. Kljub temu, da je tako velik in razširjen problem, je njeno diagnosticiranje še vedno problematično in neustrezno, kar nakazuje na potrebo po standardiziranih in bolj zanesljivih načinih diagnosticiranja depresije. Identifikacija biomarkerjev predstavlja diagnostično orodje z večjo mero objektivnosti, kar bi pomagalo izboljšati trenutno prakso, ki se zanaša na samoporočanje in klinične intervjuje. Kot tarčo za iskanje biomarkerjev v tem magistrskem delu predlagamo različne nivoje delovanja črevesnega mikrobioma. Slednji predstavlja kompleksni sistem, ki vpliva na gostiteljevo fiziologijo, metabolizem, odpornost, prehranjevanje in vedenje – mehanizmi, ki so dokazano moteni pri depresiji. V skladu s tem smo pridobili 157 sekvenciranih fekalnih vzorcev, od tega 87 bolnikov z depresijo in 70 zdravih posameznikov ter njihovih metapodatkov (starost, spol, indeks telesne mase, gastrointestinalni prehodni čas, parametri kvalitete življenja) (Projekt LifeLines). Podatke smo preprocesirali z uporabo platforme bioBakery in jih analizirali v dveh programih – Orange in JADBio. V obeh smo identificirali podmnožico biomarkerjev, ki so bili najbolj pomembni pri napovedovanju depresije. Med njimi je bil največji delež encimov, sledili so jim metabolne poti in taksoni, v JADBio pa še genske družine. Izbrane karakteristike so bile podlaga za učenje klasifikacijskih modelov za diagnosticiranje depresije v neznanih vzorcih. V obeh programih smo dobili skoraj popoln model pri razlikovanju med zdravimi posamezniki in bolniki z depresijo – v programu Orange je bil to model, ki je temeljil na algoritmu naivni Bayes (AUC 0.949), v JADBio pa Ridge logistična regresija (AUC 0.967). S tem smo pokazali, da je črevesni mikrobiom res lahko primerna tarča za iskanje biomarkerjev za depresijo in da lahko na podlagi identificiranih biomarkerjev zgradimo model, ki z visoko natančnostjo razlikuje med zdravimi posamezniki in bolniki z depresijo, kar ima ogromen potencial za reševanje problemov, s katerimi se kliniki soočajo pri diagnosticiranju depresije, še posebej v povezavi z drugimi boleznimi, ki jih lahko detektiramo preko mikrobioma.

• Language: Slovenian
• Keywords: depresija, črevesni mikrobiom, biomarker, napovedovanje bolezni, strojno učenje
• Work type: Master's thesis/paper
• Typology: 2.09 - Master's Thesis
• Organization: PEF - Faculty of Education
• Place of publishing: Ljubljana
• Publisher: M. Primožič
• Year: 2024
• Number of pages: 72 str., 5 str. pril.
• PID: 20.500.12556/RUL-160836-6c1dbde5-ab39-cae2-feea-22decd515c2a
• UDC: 616.895.4(043.2)
• COBISS.SI-ID: 206602755
• Publication date in RUL: 05.09.2024

Secondary language

• Language: English
• Title: Human gut microbiome biomarkers for prediction of depression

Abstract

Depression is one of the world's leading mental disorders and has a range of negative consequences for the individual and society. Despite being such a major and widespread problem, its diagnosis is still problematic and inadequate, suggesting the need for standardised and more reliable ways of diagnosing depression. The identification of biomarkers represents a diagnostic tool with a greater degree of objectivity, which would help improve current practice, that relies on self-report and clinical interviews. In this thesis, we propose different levels of gut microbiome function as targets for biomarker discovery. The latter represents a complex system that influences host physiology, metabolism, immunity, nutrition and behaviour - mechanisms that have been shown to be disrupted in depression. Accordingly, we obtained 157 sequenced faecal samples, 87 from depressed patients and 70 from healthy individuals, as well as their metadata (age, sex, body mass index, gastrointestinal transit time, quality of life parameters) (LifeLines Project). The data were preprocessed using the bioBakery platform and analysed in two software programmes - Orange and JADBio. In both, we identified a subset of biomarkers that were most relevant in predicting depression. Among these, enzymes accounted for the largest proportion, followed by metabolic pathways and taxa, as well as gene families in JADBio. The selected characteristics were the basis for teaching classification models to diagnose depression in unknown samples. In both programmes, we obtained a near-perfect model in discriminating between healthy individuals and patients with depression - in Orange it was a model based on a Naive Bayes algorithm (AUC 0.949), and in JADBio it was a Ridge logistic regression model (AUC 0.967). We have thus shown that the gut microbiome can indeed be a suitable target to search for biomarkers of depression and that, based on the identified biomarkers, we can build a model that discriminates with high accuracy between healthy individuals and patients with depression, which has a huge potential to solve the problems clinicians face when diagnosing depression, especially in relation to other diseases that can be easily detected via the microbiome.

• Keywords: depression, gut microbiome, biomarker, disease prediction, machine learning