″ the team reprogrammed lymphocytes (immune cells) from six entirely new bipolar patients, some of whom are known lithium responders. The team found the same hyperexcitability in the lymphocyte-derived neurons
“Although responders and nonresponders both produce more electrical impulses and spontaneous activity, when we look at the electrophysiological properties, the two groups are very different from each other.”
The Salk team characterized the electrical firing patterns of all six patients’ neuronal lines, measuring spike height, spike width, the threshold for evoking a reaction and other qualities. The overall patterns were noticeably different in responders versus nonresponders.
And machine learning :
“Wondering whether the differences could be predictive, the team trained a computer program to recognize the variations between the profiles of responders and nonresponders using the firing patterns of 450 total neurons over six independent training rounds. In each round, they started fresh with the neurons of five of the patients to train the system. They then tested the system with the neurons of the sixth patient, whose lithium status was known to the team but not to the program. They repeated the process five more times, which allowed them to build essentially six independent models. Each model was trained on the data from five out of the six patients, leaving a different patient out of the training data each time, and then letting the model classify this remaining patient as a responder or nonresponder. Using the firing patterns of just five of any patient’s neurons, the system identified the person as a responder or nonresponder with 92 percent accuracy.”
″ the team reprogrammed lymphocytes (immune cells) from six entirely new bipolar patients, some of whom are known lithium responders. The team found the same hyperexcitability in the lymphocyte-derived neurons
“Although responders and nonresponders both produce more electrical impulses and spontaneous activity, when we look at the electrophysiological properties, the two groups are very different from each other.”
The Salk team characterized the electrical firing patterns of all six patients’ neuronal lines, measuring spike height, spike width, the threshold for evoking a reaction and other qualities. The overall patterns were noticeably different in responders versus nonresponders.
And machine learning :
“Wondering whether the differences could be predictive, the team trained a computer program to recognize the variations between the profiles of responders and nonresponders using the firing patterns of 450 total neurons over six independent training rounds. In each round, they started fresh with the neurons of five of the patients to train the system. They then tested the system with the neurons of the sixth patient, whose lithium status was known to the team but not to the program. They repeated the process five more times, which allowed them to build essentially six independent models. Each model was trained on the data from five out of the six patients, leaving a different patient out of the training data each time, and then letting the model classify this remaining patient as a responder or nonresponder. Using the firing patterns of just five of any patient’s neurons, the system identified the person as a responder or nonresponder with 92 percent accuracy.”
incredibly limited data set, but a good start!