Will AI make plant science study or research any different?

By Asif Ali

Artificial intelligence (AI) is becoming increasingly popular with the advent of its feature tools for research and study. These tools have the potential to make complex things simpler and stipulated time shorter. Integration of AI into plant science could help us to understand, interpret and manipulate complex biological processes with more ease and efficiency.

Use of AI tools to accelerate plant sciences research or study

In recent years, AI tools have been extensively based on its generation of text, text to speech and text to graphics.  The future AI tools will be greener, changing the way we study and research plants in all environments ranging from dry labs (simple and complex computer-based data analysis) to well-controlled lab conditions, semi-controlled greenhouse conditions, and highly variable field conditions (Figure 1).

Figure 1: Wordcloud showing the use of AI in Plant sciences Research and Study

For data analysis, Van Noorden and Perkel surveyed 1600 researchers and found that AI tools were utilized by most of the researchers to generate video graphic content, write codes, brainstorm ideas, research manuscripts, literature surveys and in the writing of emails and grants (Van Noorden and Perkel, 2023). Several AI-based web tools have been used to accelerate research content in different ways e.g., Canva and Biorender for creating graphical content, Draw.io to generate pathways, Replit to generate codes faster than humans, Slides AI to create PowerPoint presentations, Wordtune to find word synonyms, and Listening.io to listen to the text of research articles. Similarly, several other tools e.g., Citation Gecko (citationgecko.com) to search for citations network of connected articles, Inciteful (inciteful.xyz) to connect similar articles, Research rabbit (researchrabbitapp.com) to map literature citations and Carrot2 (search.carrot2.org) to organize search results into a topic. 

AI has an impressive capability to extract data from complex tables and figures and can be used to perform meta-analysis. Several deep learning and AI-driven tools have been utilized to study organelle genomics. For instance, to study the global alignment of whole chloroplast (cp) genomes, mVISTA web-based platform can be used to generate DNA alignment (Mayor et al., 2000). In addition, for visualizing the circular cp genome, the web tool Chloroplot (https://irscope.shinyapps.io/Chloroplot/) has become well-known and can directly work with the given NCBI accession number or the raw data (.gb) file (Zheng et al., 2020). A researcher Fardous Mohammad Safiul Azam from Neijiang Normal University, who studies the chloroplast genome says that AI can be useful in the future in terms of dealing with big data generated from genomics and evolutionary studies. The scale of creating data has become tremendous in this decade. There is an obvious looming need to deal with the growing and unprecedented volume of data.

Under laboratory conditions, AI has been used in tissue culture to optimize growth conditions and help to increase transformation efficiency. In a greenhouse, AI is helpful in speedy drug discovery, as it has the potential to extract the features of plant-based medicines. Similarly, AI neural networks have been used in the quantification of phosphorus contents from leaves of strawberry, sugar beet and celery at different stages of plants (Siedliska et al., 2021).

Equally under field conditions, Hayes et al. reported that AI tools can be used to set up breeding goals and interpret genetic relationships among complex traits. Effective phenotyping of target traits and selection of parents will help to breed varieties with desired traits and will ultimately help to feed the world (Hayes et al., 2023). Integration of  AI with Unmanned Aerial Vehicles made easier phenotyping of plants, for example, it has been successfully used in the estimation of above-ground biomass in rice (Colorado et al., 2020), counting of maize tassel (Zou et al., 2020) and in the identification of different disease symptoms and their severity (Boulent et al., 2020). Drone-based remote sensing helped to study environmental variations due to the effect of climate change, for example D’Odorico studied the photosynthesis phenology of seedlings and carotenoids in conifers (D’Odorico et al., 2020).

Are we compromising the data/research integrity in adopting AI?

Despite the huge potential of AI, many researchers, especially those who had never used AI before, think that AI misuse and erroneous data will bring dishonesty in research (Owens, 2023). A respondent Sanas Mir‑Bashiri, who replied to the survey conducted by Owens said that ″ChatGPT has created a list of fictitious literature, which even do not exist″ (Owens, 2023). Similarly, one respondent Johannes Niskanen, who participated in the survey of Van Noorden and Perkel said that the Use of AI will lead us to move from ″for humans by humans″ to ″for machines by machines″ and he also indicated that ChatGPT has accumulated bad habits of humans writing e.g. use of too many irrelevant words to convey less. Scientists suggest that to avoid challenges associated with AI, it should only be used to automate the process but not to replace humans completely.

Moreover, there should be effective ways to screen the use of AI tools in prohibited situations. For example, ChatGPT itself can be used to ask if it has generated any particular text. On a safer note, it can be suggested that AI should not replace humans but rather complement humans to facilitate processing and analysis to bring pace to science.

Are we conservative in adopting AI tools?

Despite the risk of errors and misuse, AI writing assistants proved useful in writing research and reviewing manuscripts, especially for whom English is not their first language. While asking about the use of AI I felt that we are a bit hesitant and conservative about using AI in research and academia as well as day-to-day tasks. For example, not using an AI tool is similar to not using a calculator for doing complex multiplication and division, rather than sticking to by-hand calculations. It would be of course an advantage to make use of machines and tools that were developed to assist humans. On the other hand, it is also equally or even more important, and it doesn’t imply stopping learning e.g., how to do manual calculations. It would be appropriate to use AI tools while having a complete understanding of where they are impermissible (for example in drafting manuscripts) and where they are permissible (for example in writing a video lecture script). Benefits can be gained as long as generated responses are monitored by self-gained knowledge and expertise. Ideally, humans and AI will be able to find a middle ground: a human monitoring AI will stop errors in data, keep a human element in the process, and hopefully bring more benefits rather than issues.

References:

1. Boulent J, St-Charles P-L, Foucher S, Théau J. 2020. Automatic detection of flavescence dorée symptoms across white grapevine varieties using deep learning. Frontiers in Artificial Intelligence 3, 564878.

2. Colorado JD, Calderon F, Mendez D, Petro E, Rojas JP, Correa ES, Mondragon IF, Rebolledo MC, Jaramillo-Botero A. 2020. A novel NIR-image segmentation method for the precise estimation of above-ground biomass in rice crops. PloS one 15, e0239591.

3. D’Odorico P, Besik A, Wong CY, Isabel N, Ensminger I. 2020. High‐throughput drone‐based remote sensing reliably tracks phenology in thousands of conifer seedlings. New Phytologist 226, 1667-1681.

4. Hayes BJ, Chen C, Powell O, Dinglasan E, Villiers K, Kemper KE, Hickey LT. 2023. Advancing artificial intelligence to help feed the world. Nature Biotechnology 41, 1188-1189.

5. Mayor C, Brudno M, Schwartz JR, Poliakov A, Rubin EM, Frazer KA, Pachter LS, Dubchak I. 2000. VISTA: visualizing global DNA sequence alignments of arbitrary length. Bioinformatics 16, 1046-1047.

6. Owens B. 2023. How Nature readers are using ChatGPT. Nature.

7. Siedliska A, Baranowski P, Pastuszka-Woźniak J, Zubik M, Krzyszczak J. 2021. Identification of plant leaf phosphorus content at different growth stages based on hyperspectral reflectance. BMC plant biology 21, 1-17.

8. Van Noorden R, Perkel JM. 2023. AI and science: what 1,600 researchers think. Nature 621, 672-675.

9. Zheng S, Poczai P, Hyvönen J, Tang J, Amiryousefi A. 2020. Chloroplot: an online program for the versatile plotting of organelle genomes. Frontiers in Genetics 11, 576124.

10. Zou H, Lu H, Li Y, Liu L, Cao Z. 2020. Maize tassels detection: a benchmark of the state of the art. Plant Methods 16, 108.