Amyotrophic Lateral Sclerosis (ALS) is a fatal neuro degenerative disease that leads to a progressive loss of motor neurons. Despite extensive studies over several decades, ALS remains incurable with treatment only extending life by a few years. Only around 5 to 10 percent of ALS patients have a known family history of ALS (fALS) making the remainder sporadic (sALS). While fALS has been well studied using mouse models, studies on sALS have been limited since it cannot be recreated in mice. Neurons derived from human induced pluripotent stem cells (iPSC) provide a platform to study both sALS and fALS by mimicking the structure of neurons. In this work, we study electrophysiological data (Patch clamp recordings representing the ion channel behaviour of single cells and multi-electrode array (MEA) recordings measuring the extracellular activity of a population of neurons) taken from neuronal cultures derived from sALS patients and healthy controls. They are studied to gain a detailed picture of the disease mechanisms. We use machine learning, with a strong focus on interpretability to not only gain predictive insights but explanatory insights on the underlying pathology and disease phenotypes. Models are trained to classify a given patch clamp or MEA signal as to whether it corresponds to an ALS patient while interpretability methods discover the indicators that led to this classification, potentially revealing phenotypes of ALS. Additionally, we propose a new pre-processing workflow for these data to improve the classifier performance. Preliminary results indicate that supervised machine learning models can identify ALS to a reasonable accuracy while obtaining post-hoc explanations on these models can provide feature-based explanations to these classifications. Continuation of this work would identify regions of interest in the signals which are helpful to the classification and to improve model explanations, shedding light onto the drivers of ALS.