Retinoblastoma is a rare cancer of the eye that occurs primarily in young children. Studying the disease has shed light on how malfunctioning genes can lead to cancer. Retinoblastoma arises from mutations in a tumour suppressor gene that regulates cell proliferations. We each have two copies this gene – one from each parent. Thus, having a single faulty copy has no effect, either by spontaneous mutations or being inherited from a parent. If, however, both copies are mutated in the same cell, it will likely turn into a tumour. This ‘two-hit theory’ applies to tumour suppressor genes in general.

Studies of vision in humans have contributed to the development of computer vision and vice versa. Studying how artificial neural networks solve vision problems can provide insights into how biological neural networks work and why the brain computes as it does. In the 1960s, computer scientists thought they could solve the ‘vision problem’ with several months of work. It turns out that mimicking millions of years of evolution requires much more time and effort. Recent advances in technologies allow to record the activity of many thousands of neurons simultaneously on a sub-millisecond timescale. Machine learning provides methods to analyse such big data to help scientists better understand how neurons work together to process visual information and eventually generate an animal’s behaviour.

To understand the evolution of the visual system, scientists study a range of organisms that share evolutionary ancestors with humans. The evolution of colour vision, for example, is tightly linked to the molecular evolution of the opsin gene family. Animals gained or lost the ability to discriminate colour through a duplication or loss of opsin genes in evolution, respectively. The colour spectrum an animal can see is dependent on the number of opsins. We humans have three different opsins – red, green and blue – while most mammals only have two opsins. In contrast, mantis shrimps have 12 opsins, allowing them to detect even ultraviolet and polarised light.

Understanding visual information processing in the brain and replicating it in digital systems will provide key insights into developing next-generation Artificial Intelligence. It will also lead to new treatments for visual impairments, such as a visual prosthesis that can faithfully emulate the function of the eye. Implants that interface with nerve cells will then be placed into the eye or the brain to restore functional vision in blindness.