The cardiac pacemaker activity is formed from multiple interlocking physiological networks, any one of which can generate rhythm. The interlocking is reciprocal so that they automatically replace each other. In such interlocking control systems, the association scores for individual components are necessarily low, even though causation, measured by the electric current carried by the relevant ion channels, is large. This kind of reciprocally based robustness is widespread in living organisms, which explains why most association scores in genome-wide association studies are low, or even zero. It also explains why the polygenic scores do not reliably predict disease states. Genomics alone is therefore a poor method for the discovery of treatments for polygenic diseases. Reliance on genomics has led to a gene-centric impasse in healthcare, which requires a shift in favour of physiological studies that can reveal genuine causation rather than just association. The case for such a shift includes understanding that DNA is not a good self-replicator in very large genomes. Self-replication of DNA and RNA in purely chemical environments confirms that fact. The error rate would amount to hundreds of thousands in a genome of three billion base pairs. Living cells can orchestrate the enzymes necessary for nearly perfect replication before cell division by correcting those errors. These cellular proof-reading processes also open the way to control processes that are used to generate new DNA sequences when the immune and other systems need to do so.