Scite Assistant (AI) answers the question "How much of the human genome consist of junk DNA?"

Scite Assistant is billed as "your AI research partner" and as "ChatGPT for researchers." It's supposed to draw on peer-reviewed published scientific papers for its information and it will give you an answer with genuine citations.

That sounds like a good idea until you realize that the scientific literature is full of misinformation and conflicting information. What we need is an AI assistant that can help us sort throught the misinformation and give us a genuine well-informed answer on controversial issues.

Let's pick the question of junk DNA as a completley random (!) example of such an issue. The scientific literature is full of false information about the origin of the term "junk DNA" and what it was originally intended to describe. It's also full of false information about recent results and how they pertain to junk DNA.

I suspected that Scite Assistant would just find that false information and spew it back at me as an authorative answer based on the scientific literature but I decided to find out if that were true so I asked it, "How much of the human genome consist of junk DNA?"

See the answer below. No surprises. This version of artificial intelligence is no more capable of finding the correct answer than ChatGPT. [ChatGPT lies about junk DNA] The only difference is that, unlike ChatGPT, it doesn't make up references or quotations. [ChatGPT is still making up quotations from scientists] It's a lot worse than Copilot which relies heavilty on Wikipedia. [Copilot answers the question, "What is junk DNA?"]

The concept of "junk DNA" has evolved significantly since its inception, reflecting ongoing debates in the scientific community regarding the functional significance of non-coding regions of the human genome. Historically, it was posited that approximately 98% of the human genome consists of non-coding DNA, often referred to as "junk DNA" due to its perceived lack of function (Li, 2016). However, the ENCODE project, launched in the early 2000s, claimed that around 80% of the human genome is biochemically functional, a statement that has sparked considerable controversy and critique regarding the criteria used to define "functionality" (Doolittle, 2013; Palazzo & Gregory, 2014). Critics argue that the methods employed by ENCODE may overestimate the functional elements within the genome, leading to confusion about the true nature of non-coding DNA (Graur et al., 2013).

Recent studies indicate that while a significant portion of the human genome is indeed non-coding, this does not imply that it is devoid of function. For instance, it has been shown that non-coding DNA includes regulatory elements that play crucial roles in gene expression and cellular function (Antonellis et al., 2010). The complexity of non-coding sequences is underscored by findings that suggest up to 12% of the human genome may be under evolutionary constraint, indicating potential functional significance (Robyr et al., 2011). Moreover, the distinction between "junk DNA" and "garbage DNA" has been made, where "junk DNA" is seen as neutral in terms of evolutionary fitness, while "garbage DNA" actively detracts from it (Graur et al., 2015).

The ENCODE project's findings have been both supported and challenged. While some researchers assert that a large portion of non-coding DNA is functional, others maintain that many of these regions may not contribute to the organism's fitness and could be classified as "junk" (Fagundes et al., 2022; Sundaram et al., 2014). For example, it has been estimated that only about 20% of the human genome is under selective pressure, suggesting that a substantial amount of non-coding DNA could indeed be non-functional or "junk" (Kellis et al., 2014). Furthermore, ongoing research into transposable elements and their roles in genome evolution suggests that these regions, often considered junk, may have more complex functions than previously understood (Sundaram et al., 2014).

In summary, while the notion of "junk DNA" persists, it is increasingly recognized that non-coding regions of the genome may harbor significant regulatory functions, challenging the simplistic categorization of these sequences as entirely non-functional. The debate continues, with ongoing research aiming to clarify the roles of these enigmatic portions of the genome.

I decided to ask a followup question to see what references Scite Assistant identified in anwering the question, "How do you define junk DNA?"

The term "junk DNA" has historically been used to describe portions of the genome that do not code for proteins and were thought to lack any functional significance. This concept emerged in the early days of genomic research, particularly following the Human Genome Project, which revealed that only about 1-2% of the human genome is composed of protein-coding sequences (Mathews (2017); Barrett, 2009). Consequently, the remaining non-coding DNA was labeled as "junk," implying that it was largely irrelevant to biological processes (Johnson, 2014).

However, the understanding of junk DNA has evolved significantly over the past two decades. Recent research has highlighted that many non-coding regions of DNA, previously dismissed as junk, may actually play critical roles in regulating gene expression and maintaining genomic integrity (Gryder et al., 2021; Fritz et al., 2019). For instance, transposable elements (TEs), once considered mere remnants of evolutionary history, are now recognized as important contributors to the regulatory landscape of the genome, acting as sources of diverse cis-regulatory sequences (Sundaram & Wysocka, 2020). This shift in perspective is largely attributed to findings from large-scale projects like ENCODE, which suggested that a substantial portion of non-coding DNA is biochemically active and may have functional roles (Gryder et al., 2021).

Despite these advancements, the definition of junk DNA remains contentious. Some researchers argue that junk DNA should be defined strictly as sequences that do not contribute to the organism's fitness or functionality at any level (Graur et al., 2015; Doolittle & Brunet, 2017). Others propose a more nuanced view, suggesting that while some non-coding DNA may not have a direct functional role, it could still influence evolutionary processes or serve as a reservoir for genetic variation (Fagundes et al., 2022; Havstad & Palazzo, 2022). This perspective aligns with the notion that junk DNA may not be entirely devoid of significance; rather, it could play indirect roles in shaping the genome's evolutionary trajectory (Díaz-Castillo, 2017).

In summary, while the traditional view of junk DNA as non-functional remains prevalent, ongoing research continues to uncover the complexities and potential roles of these sequences. The evolving definitions reflect a growing recognition that what was once considered "junk" may indeed have important biological implications, necessitating a reevaluation of how we classify and understand these genomic regions.

Two of those citations are interesting. The Mathews (20017) refernce is a review of Nessa Carey's book Junk DNA where her definition of junk DNA is quoted. This is not a peer-reviewed paper from the scientific literature.

The Havstad & Palazzo (2022) paper is cited as an example of a possible role for non-coding DNA thus casting doubt on whether it is junk. I think this is a correct interpretation of that paper but my colleague, Alex Palazzo, claims that it was not their intent to cast doubt on the concept of junk DNA.

I should note that these answers are not reproducible. If you ask the same question again you get slightly different answers, some of which are closer to the truth while others are even more wrong.

---