In Blog

Can AI algorithms are not ready to replace science editors?

Anyone familiar with cell culture will instantly recognise the issue in this excerpt. “Roswell Park Memorial Institute 1640” (commonly “RPMI 1640”) refers to a cell culture medium named after the institute where it was developed in the 1960s. It’s highly unlikely that researchers based in Japan cultured their cells there!

So, how did such a statement find its way into the Methods section of a research manuscript? Well, this is just one of several glaring errors introduced by an artificial intelligence (AI)-based editing tool we tested.

AI: A Buzzword with Big Promises

AI is certainly the buzzword of the year, making its mark across sectors with promises of transformative potential. For scientists, one of the most exciting prospects is AI’s ability to extract meaningful insights from big datasets—a capability poised to revolutionise healthcare in the coming years. However, AI currently risks overpromising, with developers claiming their algorithms can handle tasks they’re not yet ready for, at least in untrained hands. 

Exploring AI Tools for Scientific Editing

Over the past few months, I trialled several AI applications that made bold claims in the editing arena—perfect grammar, optimised sentences, native-level English, and even advanced features like paraphrasing, plagiarism prevention and editorial reports. These features are particularly appealing given that most scientific articles (up to 98%) are published in English, despite fewer than 20% of the global population being native speakers.

To fully explore these tools, I had to commit to paid subscriptions, as free versions were significantly limited. I approached this investment with optimism, hoping the results would justify the cost. To ensure fairness and avoid endorsing or discouraging specific tools, I’ve anonymised the names of the AI services tested. If you’re interested in learning more about any of them, feel free to reach out privately for details. Read on for an overview.

Anonymised AI Tools: The Good, The Bad, and The Tedious

Tool 1: Initially promising with extra features like editorial reports, but these turned out to be basic grammatical summaries, tediously listed one by one. Its MS Word add-in required manually accepting each change, massively slowing the process and limiting its utility.

Tools 2 and 3: Provided tracked changes in MS Word, making them more user-friendly than Tool 1. However, they introduced conceptual and grammatical errors, didn’t refine text for wordiness, and failed to follow conventions of scientific writing in the biomedical sciences. Overall, a disappointment.

Tool 4: The simplest of all, delivering quick edits to a document of thousands of words in under a minute, but limited to few, minor grammatical tweaks without addressing substantial issues.

Tool 5: A hybrid of AI and human editing at a premium cost. This one I thought had the potential to be a winner. Unfortunately, the human editing offered little beyond basic grammar corrections and even introduced errors without adding comments to flag unresolved issues. For example:

In this excerpt, edits were made without any accompanying comments, implying the final text was correct. However, the edited sentence remains grammatically incorrect, poorly constructed, and unclear in meaning. 

The Risks of Over-Reliance on AI

These experiences highlight the need for caution when using AI for editing, as it seems that even costly premium services often fall short of the nuanced demands of scientific writing. While AI tools promise polished grammar and language, I found that they mainly address surface-level issues like punctuation or typographical errors, akin to standard grammar checkers, and require manual oversight to ensure accuracy. Errors introduced by AI—or even human editors—can critically alter meaning, underscoring the importance of thorough review and sound judgement in scientific editing. For example:

Without reading the edited text with the tracked changes showing, you might not readily spot this error in meaning. The authors intended to say that Telmisartan is a partial PPAR-γ agonist, but the text has been edited to say that Telmisartan was chosen partly because it is a PPAY- γ agonist. In this case, the sentence would be more accurate if it had been left alone!

Evolving but Not Yet Ready

So what’s my take home message? AI algorithms for editing scientific content are still evolving. While they show great potential for improving grammar and language accuracy, and many are financially accessible, the tools I tested are far from perfect. Polished grammar, while important, is rarely the deciding factor in effectively pitching your manuscript or funding proposal. Success lies in how you present and frame your findings—with accuracy, clarity, and a compelling narrative.

If you do choose to use AI for editing or writing, proceed with caution and thoroughly review the results. Additionally, when using AI generative tools, remember that the phrasing of your commands significantly influences the output. “Talking” to AI effectively is a skill that requires practice and precision.

Disclaimer: The views and opinions expressed in this article are those of the author and do not necessarily reflect the policies or positions of any company, organisation, employer, or other entity. The content is provided for informational purposes only and is based on the author’s personal experiences and opinions. The author makes no representations regarding the accuracy or completeness of the information presented or linked to in this article and will not be liable for any errors, omissions, or availability of the information. The author assumes no responsibility for any losses, injuries, or damages resulting from the use or display of this information.
Conflict of Interest Statement: The author of this article is an Editor for Insight Editing London, an editing services company. Every effort has been made to provide a fair and unbiased overview of other editing services; however, readers should be aware of this potential conflict of interest. The opinions expressed are solely those of the author and do not necessarily reflect the views of Insight Editing London. The author receives no direct financial benefit from discussing other editing services in this article.

In Client successes

A new study by Dandan Pi and colleagues at Chongqing Medical University unveils the role of dysregulated RAS in pediatric sepsis

In July this year, Dandan Pi and colleagues published their study on the role of the renin-angiotensin system (RAS) in pediatric sepsis in the leading journal Shock.

Pediatric sepsis is a life-threatening condition where the body’s response to infection leads to widespread inflammation and organ dysfunction. The RAS, a critical hormone system regulating blood pressure and fluid balance, is often disrupted during sepsis, worsening its severity by contributing to vascular instability, impaired immune response, and reduced organ perfusion. This study provides new insights into how RAS dysfunction influences the progression of sepsis in children, with potential implications for developing targeted treatment strategies.

Key findings include:

  1. Increased Renin Levels: Serum renin concentrations were significantly higher in sepsis patients compared to healthy controls.
  2. Decreased Angiotensin (1-7) Levels: Patients with sepsis exhibited lower levels of angiotensin (1-7), suggesting its potential as a biomarker for sepsis severity.
  3. Predictive Value: The combination of serum renin, angiotensin (1-7), and procalcitonin resulted in a diagnostic model with an AUROC of 0.87, indicating strong predictive accuracy for patient outcomes.

This research highlights the potential of circulating renin and angiotensin (1-7) as valuable biomarkers for pediatric sepsis, paving the way for improved diagnostic and therapeutic approaches to manage this critical condition in vulnerable young patients.

Join us in congratulating the authors on this groundbreaking work that enhances our understanding of pediatric sepsis! 🙌

For the full study, check it out here: Shock (lww.com)

In Client successes

A systematic review and meta-analysis from Sidra Medicine researchers highlights the link between IL-33, obesity and type 2 diabetes

We are delighted to offer our congratulations to Ghalia Missous and Nicholas van Panhuys for their latest publication on the role of Interleukin-33 (IL-33) in obesity and type 2 diabetes (T2D)! 🌍✨

As we continue to face the rising prevalence of obesity and T2D worldwide, innovative research is essential for developing novel interventions. This systematic review and meta-analysis delves into IL-33 – a cytokine that plays a dual role in the body by both regulating inflammation and contributing to the immune system’s response. In recent years, it has been increasingly recognized for its involvement in metabolic diseases like obesity and T2D. Indeed, elevated levels of IL-33 have been linked to chronic low-grade inflammation, a hallmark of obesity and a driving factor in the development of insulin resistance, which underpins T2D.

This latest study, published in August this year in the American Journal of Physiology-Endocrinology and Metabolism, examined IL-33 levels across 18 studies, revealing critical insights, including the need for improved measurement methods to reduce heterogeneity in findings. These results are pivotal for enhancing our understanding of IL-33’s role in disease progression and guiding future research.

🔗 Check out the full study here: Circulating interleukin-33 levels in obesity and type 2 diabetes: A systematic review and meta-analysis | American Journal of Physiology-Endocrinology and Metabolism

Let’s celebrate this important work and encourage further exploration into the potential of IL-33 in combating obesity and T2D! 💡

In News

We were delighted to see that the latest findings from Adam Claridge Chang’s lab published in Nature Communications in April, were covered in The Straits Times just last week!

In case you missed it in April and would like to add this paper to your summer reading list, you can find the article here: https://www.nature.com/articles/s41467-024-47203-w#Sec17

So, what’s it all about?

Light-gated, chloride-conducting anion channelrhodopsins (ACRs) are powerful neuron inhibitors. The problem is that for cells with high intracellular chloride concentrations, ACRs may have an activating rather than inhibitory effect. So, Adam Claridge-Chang and colleagues aimed to find an alternative optogeneic approach to using ACRs. Enter, potassium-conducting kalium channelrhodopsins (KCRs)!

What did they find?

The research team evaluated the ability of KCRs to inhibit neural activity and suppress behavior in three model organisms: Drosophila, Caenorhabditis elegans, and zebrafish. They found that a variant of KCR1, with improved plasma-membrane trafficking, showed comparable potency to ACR1 but with enhanced properties, including reduced toxicity and better efficacy in high-chloride cells.

What does this mean?

KCRs could be considered next-generation optogenetic inhibitors, opening new avenues for in vivo circuit analysis in small, genetically tractable animals.

Congratulations to the whole team who worked on this exciting project – and for the exciting coverage in The Straits Times: https://www.straitstimes.com/singapore/duke-nus-researchers-develop-light-controlled-switch-for-brain-cells-to-better-study-the-brain

#Neuroscience #Optogenetics #ResearchInnovation #NeuralCircuits #KCRs #ACRs #InVivoAnalysis #Biotechnology

In News

We are delighted to see that the latest study on the mechanisms of chemo-resistance in T-cell Acute Lymphoblastic #Leukemia (T-ALL) by Jingliao Zhang and colleagues is now published in the prestigious journal, Blood!

T-ALL is an aggressive cancer not least because of the propagation of resistant cancer clones that drive disease recurrence. Jingliao Zhang et al. (Institute of Hematology and Blood Diseases Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College) wanted to dissect the nature of these clones to work out how their presences might contribute to resistance to #chemotherapy .

Combining single cell RNA sequencing with T-cell receptor sequencing of paired diagnosis-relapse T-ALL samples, the researchers identified two leukemic evolutionary patterns: “clonal shift” and “clonal drift”. They additionally saw high expression of the RNA-binding protein MSI2 in the clones persisting at the point of disease relapse. Digging deeper, the researchers conducted functional studies showing that MSI2 contributed to T-ALL proliferation and promoted #chemoresistance through the posttranscriptional regulation of the #oncogene, MYC.

These findings have important implications, as they identify MSI2 as an informative biomarker and novel therapeutic target in T-ALL.

Congratulations to all those involved in this intricate study! For those of you who would like to learn more, the paper can be found online here: https://lnkd.in/dcRJVgr9

In News

New insights into physiological and pathological brain wiring

We’re excited to share news of the publication of a fantastic article by an IEL client last month, edited by IEL’s Ilya Demchenko.

Published in PNAS as an open access article, Sinclair-Wilson and colleagues describe their ground-breaking work on the plasticity of brain circuits in neonates, which is important for the correction of embryonic thalamocortical axon mis-targeting. Using a genetic mouse model, the researchers identified a serotonin-dependent window in the immediate post-natal period in which pre-natal axon miswiring can be corrected and appropriate definition of cortical areas rescued: this period was disrupted by pre-term birth and dysregulation of serotonin levels. This work may have profound implications for our understanding of human neurodevelopmental disorders that occur in extremely pre-term infants.

You can find out everything you need to know by downloading the full text here: Plasticity of thalamocortical axons is regulated by serotonin levels modulated by preterm birth | PNAS

Well done to everyone involved in this groundbreaking study – it was a pleasure working with you and we look forward to learning how this work progresses in the future!