Generally, it is difficult to carry out conventional experiments on living cells to predict genes. But today, bioinformatics research is making it possible to predict the functions of a gene, based on its sequence alone. The DNA databases that are cropping up with many DNA sequences around the world facilitate the gene prediction process intensively. Have you ever wondered about Gene Prediction? It is the process of identifying potential coding regions in an uncharacterized region of the genome. Simply it means locating genes along a genome.
Prediction of genes to find the location of protein-coding regions is now an active research field in bioinformatics. Genes are not the only thing looked for in this process, but also splice sites, protein binding sites and DNA 3D structural features. Gene finding is specific to each species, therefore, functional regions may vary by species. Moreover, the common repeat sequence information is considered in gene-finding programs to identify coding regions.
Gene finding is less difficult in prokaryotic genomes due to the absence of introns in the protein-coding genes. Introns do not code for any functional protein but exist between functional regions(exons), which are usually responsible for protein synthesis. Eukaryotic gene prediction is a more complex problem as the genomes are much larger than prokaryotic genomes and also they tend to have a very low gene density. For example, in humans, only 3% of the genome codes for genes. Moreover, the identification of exons and introns has made it difficult for gene prediction.
When discussing the methods of gene prediction, similarity-based searches and ab-initio predictions are important. Sequence similarity search is based on finding similarities in gene sequences between ESTs (Expressed Sequence Tags) and proteins. When a similarity between a certain genomic region and an EST, DNA, or protein is discovered, the similarity information can be used to infer the gene structure or function of that region. Local alignment and global alignment methods are used in this type of search. The most common tool used is BLAST and it detects sequence similarity to known genes. Software such as PROCRUSTES and GeneWise use the global alignment method for gene prediction.
Ab-initio methods use gene structure as a template to detect genes and it is a computational task. Many algorithms are applied in ab-initio methods for modeling gene structure, such as Dynamic Programming, Hidden Markov Model and Neural Network. Based on these models, many gene prediction programs have been developed such as FGENESH, GeneParser, GlimmerM and GENSCAN.

Even though there are many gene prediction programs, the accuracy and reliability of the algorithms used must be considered as well when using such programs. Furthermore, comprehensive criteria are required to evaluate gene prediction programs. One of the main drawbacks is that it is still difficult to locate short exons in genes, therefore the performance of algorithms for recognizing short coding sequences is quite low. Other challenges involved in gene prediction are sequencing errors in raw DNA data, dependence on the quality of the sequence assembly, handling short reads, frameshift mutations, overlapping genes and incomplete genes.
Since the outcome of the gene prediction approach is quite essential in transcriptomics, proteomics and genome studies, it requires great effort in both computational and experimental methods to make gene prediction more accurate. This will in turn reduce the amount of experimental verification work required significantly. With rapid advances in computational techniques and understanding of the splicing mechanism, it is hoped that reliable eukaryotic gene prediction can become more feasible in the upcoming future.

References:
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5187414/
https://www.sciencedirect.com/topics/medicine-and-dentistry/gene-prediction

Image courtesy:
Featured image:
https://www.sciencemag.org/news/2018/09/there-are-about-20000-human-genes-so-why-do-scientists-only-study-small-fraction-them

Figure 1:
https://www.slideserve.com/gomer/gene-finding-and-gene-structure-prediction