Post-polyploidy subgenome evolution of Glitch Art

Can artistic innovation be inspired and guided on knowledge taken from the field of plant genome evolution? I explored this question by subjecting a glitch artwork to processes analogous to those

shaping the evolution of the maize genome, that are whole genome duplication (WGD), subgenome bias fractionation and genome dominance, respectively. The mosaic composition of glitch art provided me with a comparable visual reference to the mosaicism found in plant genomes that emerged from multiple rounds of polyploidy (genome doubling) events throughout the history of many plant species.

Mosaicism as shared characteristic of the maize genome and Glitch Art

The maize genome is a mosaic of rearranged chromosomal segments that arose as a result of allotetraploidization - the hybridization of two species followed by chromosome doubling - and subsequent diploidization - the transition of a genome from polyploid to diploid -. As a consequence, maize is an ancient tetraploid that behaves as diploid. Glitch art is a mosaic of pixels created by artists that use data compression, loss and error methodologies that highlight technology's inability to execute its intended function, producing visual noise artifacts that are considered art.

Plant scientists also explore technological means to produce 'error' and take advantage of the organismal's inability to execute its biological processes as procedural approach to generate novelty in plant form and function. For instance, the use of the chemical colchicine to create synthetic polyploids (sometimes termed neopolyploids) by impairment of microtubule formation and cell division, trigger a range of biological processes that are not directly controlled by the scientist but that arises as an intrinsic mechanism of organismal biology. Thus, glitch artists and scientists rely on induced errors to create new kinds of abstraction.

Whole genome duplication in maize occurred naturally around 4.8 million years ago. Interestingly, glitches that produce unexpected visual outcomes also occur naturally and unintended when technology fails. Here I present two glitches that spontaneously arose while working on my Apple laptop and desktop computers at the time I had several programs running in parallel that caused the system to crash and stall. Because of the appealing nature of the visual outcome, I saved the resulting images as glitch art (Figure 1 and 2). The mosaicism of these art pieces is striking.

Figure 1. Visual output of spontaneous glitch that was saved as image when I experience a system crash on my MacBook Pro (17 inch) running on OS X El Capitan in 2016. Remnants of my digital drawings intermixed with segments of several windows I had open in parallel, including a youtube video on how to use Kinect 2 library in processing, can be recognized from the glitch image.

Figure 2. Visual output of spontaneous glitch that was saved as image when I experienced a system crash on my iMac (20 inch) running on OS X El Capitan in 2017. I had several windows open at the time of the crash.

The objective of the current study was to utilize glitch artwork shown on Figure 2 and intentionally subject it to artificial processes that emulate those shaping the evolution of the maize genome in order to explore potentially new visual outcomes that could enhance the aesthetic value of the image in question.

Whole genome duplication, bias fractionation and genome dominance as raw material for evolutionary innovation

Maize and sorghum are closely related grass species that shared a common ancestor about 12 million years ago, but maize underwent an additional round of genome doubling (whole genome duplication) about 4.8 million years ago compared to sorghum. This means that for each chromosomal segment that was present in the common ancestor of sorghum and maize and is still conserved among the two species, maize have two copies of the segment compared to only one copy in sorghum (2:1 ratio). Interestingly, both copies are not always retained in the maize genome (composed of two subgenomes), with the removal of one of the copies during diploidization. The removal of duplicated gene copies in the maize genome is known as fractionation and it has been shown that this process is biased, with gene loss more frequently associated to one of the subgenome compared to the other.

Apparently, single gene loss by deletion (compared to inactivation and sequence randomization, gene transposition, and segmental transpositions) seems to be the most frequent fractionation mechanism in the maize genome. This means that genes are being deleted and lost from one of the subgenomes at a higher rate than on the other subgenome. The 'tension' of gene loss versus retention between duplicated genome segments is worth exploring from an artistic perspective.

Levels of gene expression influence differences in the rate of gene loss because the most highly expressed gene of the homeologous pair tend to be conserved. Furthermore, genes from the maize subgenome that is experiencing higher rate of fractionation are in general expressed at lower levels. This phenomena of one subgenome displaying higher expression levels than the other subgenome has been termed genome dominance. It has been proposed that purifying selection acting against deletion alleles of gene copies that contribute more to total gene pair expression is the main evolutionary force driving and maintaining bias fractionation and genome dominance processes. In this work, the process of genome dominance will also be explored from an artistic perspective.

The processes described above account in part to the great genomic and phenotypic diversity observed in maize, and has been suggested that whole genome duplication followed by bias fractionation and genome dominance are raw material for evolutionary innovation in plants. Thus, I wanted 'transfer' this process to art in order to create novelty in the visual realm of glitch art.

Whole PIXEL duplication, bias PIXEL fractionation, and PIXEL dominance as raw material for ARTISTIC innovation

In order to transfer concepts from 'maize genome evolution' to 'glitch pixel evolution' I took into account the following premises:

1 glitch artwork shown on figure 2 is the equivalent of a pre-duplication haploid genome

2 each pixel in figure 2 is the equivalent of a gene in maize

3 glitch artwork on figure 2 is subjected to whole image duplication

4 the fate of duplicate genes following whole genome duplication is equivalent to the fate of duplicate pixels following whole image duplication

5 differences in level of gene expression is equivalent to differences in pixel brightness and/or pixel color

6 gene loss after whole genome duplication is equivalent to pixels painted in black color after whole image duplication

7 assignment of duplicated segments to an ancestral and unduplicated genome is equivalent to assignment of duplicated pixels (and/or duplicated segments of pixels) to an ancestral and unduplicated image

Although the above premises can be considered to be somehow subjective, I tried to make the necessary artistic and coding decisions that could be supported on knowledge from maize genome evolution as to effectively bridge art with science. The term 'evolution' here does not imply

the use of genetic algorithms to evolve a glitch image but rather refers to the process that shaped the maize genome and how I used this process and applied it to a glitch image.

Regarding the premise #7, it is important to mention that an unduplicated image is essential for the identification of fractionated duplicate pixel segments as for differentiated lost from one duplicated subimage but retained in the other.

Overall, the process is the artistic equivalent of duplicating an artwork and subsequently deleting parts of it by painting them black, while keeping track of each pixel lost by comparing to the original unduplicated image. Pixels painted black after whole image duplication are determined by their brightness level and can be considered as pixel brightness driven bias fractionation.

Whole IMAGE duplication

Assuming figure 2 is the equivalent of an haploid genome of maize that undergoes tetraploidization, I created a single image by repeating the same glitch artwork four times (Figure 3). The top row of images will be considered as 'subimage1' and the bottom row of images will be considered 'subimage2' in relation to the two subgenomes of maize after tetraploidization (named maize1 and maize2 respectively).

Figure 3. Shown is a 'tetraploid' image composed of repeated glitch artwork previously shown on figure 2, with each visual element repeated four times.

Bias PIXEL fractionation & PIXEL dominance

The equivalence of analyzing gene expression is to look at pixel's brightness and pick those pixels higher in brightness according to a given threshold, and then 'delete' (by applying black color) all pixels with brightness level lower than the selected threshold. The result of this methodology is shown in Figure 4.

Figure 4. Pixels from figure 3 that displayed brightness levels below the threshold of 127.5 were 'deleted' by painting them with black color.

At this step, an equal threshold was applied to the two subimages and pixel loss/deletion was homogeneous across figure 4. Interestingly, equal gene loss (unbiased fractionation) and no bias in gene expression between duplicate regions (genome equivalence) is the norm for genome evolution after whole genome duplication in banana, soybean and poplar species. In order to emulate the process of bias fractionation and genome dominance occurring in maize and transfer it to figure 3 and 4, I applied a different brightness threshold to subimage2 only and kept the previous threshold for subimage1. In this manner, bias pixel deletion displayed a higher rate for subimage2 compared to subimage1 (Figure 5).

Figure 5. Differential brightness thresholds were applied to subimage1 and subimage2 that resulted in higher rate of pixel 'deletion' on subimage2. Brightness threshold for subimage2 was 212.5 (meaning all pixels with brightness below 212.5 were painted black) compared to brightness threshold of 127.5 for subimage1.

From retained pixels (pixels still present in subimage1 and subimage2) with the same brightness I extracted their amount of red color according to their position within the image along the y access. As vertical lines of pixels are defined across subimages, their color is dependent on their combined composition of red, green, blue for subimage1 and 2, giving rise different colors according to their amount of red along the Y axis. This methodology imposed expression divergence of retained pixels among subimages and was thought to emulate the phenomena known as regulatory neo-functionalization. Expression divergence and regulatory neo-functionalization would then visually translate on retained pixels among both subimages having diverged in their colors based on their amount of red along the Y axis (Figure 6). This process in maize determines that duplicated gene copies can acquire different expression patterns and possibly different function.

Figure 6. Expression divergence and regulatory neo-functionalization of pixels based on their amount of red component of color along the Y axis. Retained pixels among the two subimages will differ in color because their red component varies along the Y axis, giving rise to expression divergence (different amounts of red) and regulatory neo-functionalization (different pixel color). A vertical line every 30 pixels across the image is shown.

Syntenic alignment of retained pixels from the tetraploid image to the progenitor diploid image highlights pixel's diversity after whole image duplication

In order to analyze pixel diversity created as a result of whole image duplication, bias fractionation, pixel dominance and expression pixel divergence, I created a single image (Figure 7) that included figure 2 (progenitor diploid image), and figure 6 (diverged tetraploid image) and took a single row of pixels per diploid set: (a) one row of pixels for diploid image from figure 2, (b) one row of pixels from tetraploid subimage1, and (c) one row of pixels from tetraploid subimage2. Each row of pixels had the exact same Y value or height for the alignment to be considered syntenic. I then accessed each pixel's color data for the three rows in order to conduct the alignment (Figure 8).

Figure 7. Composite image displaying diploid progenitor image prior to whole image duplication on the top; and tetraploid image subjected to bias pixel fractionation, pixel dominance and expression pixel divergence on the bottom.

Figure 8. Corresponding row of pixels were extracted from diploid and tetraploid images and their color data was accessed in order to perform syntenic alignment of pixels at orthologous image regions. Alignment in this context means the drawing of lines joining pixels that share the same color.

By visually aligning pixels from the progenitor diploid image and compare them to the diverged tetraploid image (Figure 9 and 10) I could effectively address from a visual perspective the variation that emerged from subimage1 and 2 after processes that emulated maize genome evolution were applied to glitch artwork from figure 2. This guided and methodological approach inspired in biological principles opposes the common approach of using random functions in programming to create visual novelty purely based on computer calculations of chance. Similarly, standard processes used in glitch art weren't needed either in order to create pixel diversity within the same image.

Figure 9. Pixels from figure 8 that retained the same color were joined with a line of of equal color. Divergence of subimages after tetraploidization from the diploid progenitor created the variegated pattern without the explicit usage of random functions and or glitch procedures per se.

Figure 10. Pixels from figure 8 that retained the same color were joined with lines of while color, giving rise to a 'barcode' pattern reflecting diversity that was creates as result of tetraploidization and divergence among subimages relative to the diploid image progenitor.

Communicating science concepts in artistic form

The ability to emulate evolutionary processes shaping the genome of plants and communicate them in non-scientific form through glitch art and computer programming allows for the visual perceptualization and transmission of genomic concepts to society. Furthermore, the work presented here is an extension of the previously conceptualized artistic form of expression that I coined with the name of GAGAISMO (from Genomic and Geometric AbstractionISM) as the intersection of genomics, art and computer programming. By taking concepts from plant genome evolution as raw material for novel avenues of visuality, gives the work a conceptual framework that bridges the analytical and creative domains of human endeavors. In fact, the ideology driving the efforts of glitch artists and scientists is not that much different when it comes to coerce technological and biological systems into error creation as new forms of abstraction. In this particular case, I transited from complexity to simplicity as the end result is an image reduced to a pattern of white lines on a black background. Visual simplicity as the end result of scientific reductionism is the strength of the work presented here.

What to do with the alignment of pixels in terms of art?

I will use the concepts presented here and apply them to the creation of geometric abstract art based on pixel alignment of ancestral and duplicated images (glitch and others) as exploratory approach to algorithmic art in general and Arte GAGAISTA in particular.


Jeff Bennetzen & Sarah Hake (2009) Handbook of maize: genetics and genomics. Published by Springer Science + Business Media.

Olivier Garsmeur et al. (2013) Two evolutionary distinct classes of paleopolyploidy. Molecular Biology and Evolution, 31(2): 448-454.

Thomas E. Hughes et al. (2014) The impact of widespread regulatory neofunctionalization on homeolog gene evolution following whole-genome duplication in maize. Genome Research, 24: 1348-1355.

Wen-kai Jiang et al. (2013) Prevalent role of gene features in determining evolutionary fates of whole-genome duplication duplicated genes in flowering plants. Plant Physiology, 161: 1844-1861.

Carolyn L. Kane (2017) Exhaustion aesthetics. Leonardo, 50 (1): 5-11.

Simon Renny-Byfield et al. (2017) Gene fractionation and function in the ancient sub genomes of maize. Molecular Biology and Evolution, 0 (0): 1-8.

James C. Schnable et al. (2011) Differentiation of the maize subgenomes by genome dominance and both ancient and ongoing gene loss. PNAS, 108(10): 4069-4074.

Daniel Shiffman (2015) Learning Processing: a beginner's guide to programming images, animation, and interaction. 2nd Edition published by Morgan Kaufmann.

Jonathan F. Wendel et al. (2016) Evolution of plant genome architecture. Genome Biology, 17: 37.

Margaret R. Woodhouse et al. (2010) Following tetraploidy in maize, a short deletion mechanism removed genes preferentially from one of the two homeologs. PLoS Biology, 8 (6): e1000409.

Gregory Zinman (2015) Getting messy: chance and glitch in contemporary video art. Taken from the book 'Abstract video: the moving image in contemporary art': Edited by Gabrielle Jennings; Published by University of California Press.