The goal is to input data from a diagnosis, such as blood tests with specific markers, or use a multimodal approach that captures various aspects of a presenting phenotype. A black box model can then predict a combination of drugs and therapeutics to address the problem associated with the phenotype.
Biology aims to provide the black box with extensive, multimodal information to enhance its accuracy for each cell type in an individual, making the approach personalized. However, constructing different models for each person poses a challenge. Machine learning and deep learning models have been used to train on populations and make individual predictions, but these models often fail to generalize to individuals, relying instead on population-based predictions.
Future bioinformatics should focus on integrating all available information to predict at an individual level. Beyond prediction, the field needs to synthesize data from various modalities to create a comprehensive storyline for each patient. Viewing each patient as a time series model with synthesized data for each modality can provide a detailed narrative. This approach contrasts with targeting single genes based on presenting phenotypes, which has not achieved the desired outcomes despite numerous attempts with single-cell inference models.
Each patient requires a storyline that encompasses all body parts and the different modalities present. This comprehensive approach will improve the accuracy and effectiveness of personalized treatments.