Can TargetDiscovery really find a needle in a haystack?
First of all, TargetDiscovery can only identify a signal if it rises above the noise level. In other words, if there is not enough information in the data, no algorithm can reliably find a needle in a haystack, including TargetDiscovery. If the data contain sufficient information to differentiate signals from noise, then the state-of-the-art statistical methodology employed by TargetDiscovery gives it a better chance to pick up these signals than conventional statistical algorithms. The key features of this advanced methodology are the following: – TargetDiscovery combines the input from two separate types of filters to distinguish signal from noise. Even if one of the filters fails to correctly capture the signal, TargetDiscovery still has a chance to identify the signal through the other filter (see double robustness property). – In addition, TargetDiscovery adapts the choice of filters to the dataset at hand. It selects the filters that are best suited to capture the particular sign
