Recent breakthroughs in genetic research may have revealed new genes underlying common psychiatric disorders. Schizophrenia and bipolar disorder affect more than 64 million people around the world. These disorders are strongly influenced by genetics. However, no gene determines one’s risk of developing schizophrenia or bipolar disorder. Rather, it is likely that a number of genes contribute to the risk. Using artificial intelligence, researchers at Stanford University have now discovered complex variants throughout the human genome that may contribute to these psychiatric disorders. This new study suggests that mutations that occur after conception, such as genetic mosaicism, may be responsible for a number of psychiatric disorders including bipolar disorder and schizophrenia.

Think of a genome as a living book with instructions for every cell in the body. Our genes are the chapters. We have about 20,0000 genes that provide instructions for making proteins, the building blocks of life. However, the vast majority of our genes are non-coding, meaning they do not provide instructions for proteins. Nevertheless, these genes play an important role in genetics and regulation of cell function.

Genetic variants, or spelling changes, in either a coding or non-coding region can interfere with how the cell translates specific instructions. A small typo can have little or no effect on how the book reads. But major spelling changes can lead to the deletion of a sentence or even an entire chapter. Without the correct instructions to produce specific proteins, these spelling changes can contribute to disorders that affect various aspects of our body.

Our genes are a combination of the DNA we inherit from our parents. We have two copies of each gene, one from mom and the other from dad. These randomly sorted gene pairs determine characteristics such as hair texture, eye color and even certain health risks. Some traits are dominant, meaning that only one copy of the variant is needed for expression. Others are recessive and only show up if both copies are the same. This is called Mendelian inheritance, named after Dr. Gregor Mendes’ first observations of how genes are transferred in pea plants.

In the earliest stages of life, DNA undergoes several rounds of replication. Trillions of cell divisions occur, during which one cell divides into two identical daughter cells. However, DNA replication is prone to making mistakes. Every time a cell divides, small typos are produced in the genome. Rapid replication during the first trimester of pregnancy can therefore introduce a host of genetic changes not seen in either mother or father. This is known as genetic mosaicism, where two or more genetically distinct cell populations are expressed in the body. Mosaicism can appear as two different colored eyes, or alternating skin patterns as shown below. A number of conditions have also been associated with mosaicism such as developmental delays, autism, epilepsy and certain cancers. We all have some degree of genetic mosaicism in our bodies. This is why identical twins can have different fingerprints.

Genetic variants can also be acquired during an individual’s lifetime that further alter the mosaic of our genome. Changes in DNA can occur from exposure to chemicals or radiation, or from infections such as hepatitis B and C that corrupt the genetic material of a host cell. Other variants are acquired randomly. DNA can develop errors during replication and other normal cellular functions. This damage is exacerbated by inflammation, aging and lifestyle choices such as smoking and poor diet. Pinpointing which variants contribute to certain disorders can therefore sometimes be a very complex process.

Whole genome sequencing (WGS) can help identify small changes in DNA. This genetic test maps an individual’s entire genome using samples collected from blood or control samples. Whole genome sequencing extracts the exact sequences that make up each chapter of our DNA. The extracted sequences are then compared to reference genes from a typical human genome. Each difference between an individual’s genome and the reference genome reveals a potential variant that may be associated with a disorder.

Alexander Urban, senior author of this study and associate professor at Stanford, describes: “Looking for simple variants is like proofreading a book manuscript and looking exclusively for typos that change single letters. You overlook words that are garbled or duplicated, or in the wrong order – you might even miss that half a chapter is missing.” Some disorders may actually be linked to long, complex spelling changes in an individual’s genes, further complicated by the fact that variants across multiple genes may overlap with more than one disease.

Many psychiatric disorders are affected by multiple changes in similar genes. Bipolar disorders and schizophrenia are excellent examples of the complexity of the human genome. Hundreds of genetic variants have been identified as contributing to risk. Many of these genes are linked to brain development, regulation of the immune system and neuron signaling pathways. The AKAP11 gene in particular has been shown to be a strong risk factor for bipolar disorder, although new studies in mice suggest that this gene may also be involved in schizophrenia. Understanding how spelling changes in this gene interact with other high-risk variants may help decipher what induces the onset of psychiatric symptoms.

In their study, Zhou et. al compared the genomes of over 4,000 individuals around the world. Their entire DNA sequence was extracted using whole genome sequencing. The data was then uploaded to an AI algorithm trained to recognize dozens of genomes across different ancestries. This approach allowed researchers to match large, complex gene variants with specific health conditions.

The study specifically recruited individuals with known bipolar disorder or schizophrenia diagnoses and compared them to healthy controls. This type of approach is known as a genome-wide association study (GWAS). Genome-wide association studies compare the genes of individuals with a particular disease to a large cohort of controls. Although this approach can tell us where variants exist, this information is often not precise. For example, it can tell us that the book contains spelling changes on pages 122, 296, and 731, but not what type of error it is. The AI ​​algorithm developed by Zhou et. al adds more specificity. It highlights the changed word or sentence and reports whether it has been garbled, duplicated or deleted.

With more than 85% accuracy, the AI ​​tool identified more than 8,000 complex variants. Many of these spelling changes were found in parts of the genome that provide instructions for brain function. To determine whether these variants could be linked to psychiatric disorders, they extracted DNA from brain tissue samples from individuals affected by schizophrenia or bipolar disorder. The complex variants they identified appeared to overlap with individual variants found in other genome-wide association studies of these diseases. For example, one complex variant they found correlated with schizophrenia and bipolar disorder was 4,700 base pairs long, the basic unit of DNA. In the book analogy, base pairs are like the words in the book.

New innovations in genetic research are deepening our understanding of the human genome. By analyzing large amounts of genetic data, AI technology is revealing intricate associations between large variants and certain psychiatric disorders. This not only increases our understanding of the genetic basis of these diseases but also paves the way for personalized medicine. As we continue to uncover more of the human genome, future studies may reveal deeper insights into the genetic basis of a range of diseases.