Explainable deep learning for rapid biomarker detection

Our study explored how advanced machine learning techniques can improve the detection of biomarkers using surface-enhanced Raman spectroscopy (SERS). SERS is a fast and cost-effective method for analyzing biological samples, but interpreting its data accurately has been challenging.

We developed a three-step framework involving noise reduction, precise biomarker quantification, and a novel explainability method called CRIME. This framework allowed us to identify serotonin levels in urine with high accuracy, while also explaining how the model reached its predictions. These improvements could make SERS a powerful tool for discovering new biomarkers and diagnosing diseases quickly and affordably.

Repurposing Existing Drugs for Schizophrenia Treatment

Our review examined how drug repurposing—finding new uses for existing medications—can accelerate the development of novel therapies with fewer side effects. We analyzed 61 clinical trials conducted between 2018 and 2024, identifying 40 repurposed drug candidates targeting diverse mechanisms such as immune modulation, neurotransmitter balance, and metabolic regulation. 

By leveraging biomarker-driven patient selection and in silico drug discovery methods, the evolving landscape of schizophrenia treatment highlights the potential for more personalized and effective therapeutic strategies.

The impact of genetics on urinary metabolite concentrations

Our research looked at how genetics may influence levels of small molecules in urine, which can reflect various body functions and potentially signal health or disease states. By analyzing data from several genetic studies, we identified connections between specific genes and the levels of certain urine metabolites, including three new genetic links with tyrosine, glycine, and 3-aminoisobutyrate. These findings help us understand how our genes impact metabolite regulation and could pave the way for using these genetic markers to assess health or risk of disease in a more personalized way.

Cell biomarkers predicting statin induced cognitive decline

Our research investigated whether simvastatin, a drug commonly used to lower cholesterol, could also enhance cognitive function in schizophrenia patients. 

By examining blood markers before treatment, we identified that changes in insulin receptor levels on B cells could predict cognitive changes after one year of simvastatin treatment. This suggests a step towards personalized medicine in schizophrenia, using a patient's unique biological markers to determine the most effective treatment. Further research could highlight whether the result could be generalized to healthy statin users as well.

Urine metabolites could predict psychiatric disorders based on genetic evidence

In our study, we explored early detection of psychiatric disorders through urinary metabolites. Using genetic methods, we found distinct urinary markers for several conditions: tyrosine linked to schizophrenia, creatine to bipolar disorder, and specific markers like pyridoxal and ferulic acid 4-sulfate for anorexia nervosa, as well as N,N-dimethylglycine for ADHD.

These insights could lead to a future diagnostic method based on these markers, with the potential to accurately distinguish between various mental health conditions. This development promises earlier, more precise diagnoses, paving the way for improved treatments and outcomes in mental health care.

Machine learning model for schizophrenia diagnostics using blood biomarker data

In this study, we were trying to find a way to detect schizophrenia early by looking at white blood cells. We found that people with schizophrenia have different levels of certain substances in their blood cells, specifically more of some (called IR and CD36) and less of another (GLUT1). 

We then created a method to diagnose schizophrenia based on these blood markers. In our tests, this method was fairly accurate and could tell apart not only people with schizophrenia from those without it but also people with schizophrenia from those with other mental conditions like bipolar disorder or depression. These findings are exciting because they could help doctors diagnose schizophrenia more quickly and accurately, leading to better treatment and recovery.