Wednesday, April 24, 2024

AI Corner: SCITE.AI Auto-Generates Short Medical Articles with Clear Citations

A year ago, the bad news around any AI-generated writing was hallucinations (such as made-up bibliography.)   The field moves fast and some sources seem to have completely licked that problem.

I sometime use SCITE.AI (subscription) which has two main modes.  

In the first mode for SCITE, you search or input a key paper ("key paper.")   SCITE instantly gives you all the papers that have cited it, AND, if full text is available, the paragraph or sentences that cite the key paper.   If the key paper has been cited 15 times, at a glance you can see how the key paper was cited, rather than you go and download and/or buy all 15 of the PDFs and read them one by one.

But here's a second mode for SCITE.   It's called SCITE Assistant.   You can give it a targeted question or request, and it instantly provides a correctly footnoted essay in response.

For example, I gave it this question:

  • "How important is minimal residual disease molecular testing today, in oncology?"

Here is the automated answer:


Minimal residual disease (MRD) molecular testing is a crucial aspect of oncology today. MRD refers to the small number of cancer cells that persist in a patient after treatment. It is a significant prognostic factor in various types of cancer, including acute lymphoblastic leukemia (ALL) (Tierens et al., 2021; Pui et al., 2015; Brüggemann & Kotrová, 2017). Detecting MRD is essential for identifying patients at high risk of relapse and shorter survival, aiding in refining risk stratification and treatment response assessment (Ravandi et al., 2018; Pui et al., 2015). 

Advancements in molecular testing techniques such as real-time quantitative polymerase chain reaction, digital polymerase chain reaction, and next-generation sequencing have enabled the detection of MRD at levels far below what is required for morphological remission (Ball & Stein, 2019). These techniques have revolutionized the field by providing highly sensitive methods to monitor MRD and predict the risk of relapse in cancer patients (Johnston et al., 2023). 

MRD testing is not limited to hematologic malignancies but extends to solid tumors as well. Studies have shown that assessing MRD in solid tumors post-surgery or treatment can guide clinical decisions related to adjuvant therapy, emphasizing the importance of MRD detection in improving patient outcomes (Larribere & Martens, 2021). 

Furthermore, MRD monitoring is essential in acute myeloid leukemia (AML), where it serves as a critical biomarker for prognostic, predictive, monitoring, and efficacy-response assessments (Heuser et al., 2021). The ability to quantify MRD using various molecular techniques, including whole-genome sequencing and immunoglobulin clonality assays, has become integral in tailoring treatment strategies for patients with leukemia and other cancers (Subhash et al., 2021; Lee et al., 2022; Fürstenau et al., 2019).

 In conclusion, MRD molecular testing plays a vital role in oncology today by providing clinicians with valuable information to guide treatment decisions, predict relapse risk, and improve patient outcomes across a spectrum of cancers.  [END SCITE AI WRITING]

Then, it provides a listing of all the papers it cited, AND, for each citation, it clips several sentences to show you what facts it was pulling from that citation.

Here's the top part of the footnotes section, which explains why it chose Tierens et al. as the first reference in the AI essay above.

[SCITE.AI Continues]

Reference #1 Tierens et al

“…Measurable (minimal) residual disease (MRD) is an established, key prognostic factor in adult B-cell acute lymphoblastic leukemia (B-ALL), and testing for MRD is known to be an important tool to help guide treatment decisions.…”

[click] See full context
Section quoted: Abstract

Consensus Recommendations for MRD Testing in Adult B-Cell Acute Lymphoblastic Leukemia in Ontario

Anne Tierens1, 
Tracy L. Stockley2, 
Clinton J. V. Campbell3 
et al. 2021
Journal: Current Oncology

8       0  7  0
The paper (Tierens) has been cited 8 times.   Of these, 7 have full text.    Of the 7 with full text, 0 "support" the Tierens, 7 "mention" Tierens, 0 "disagree" with Tierens.

[Continues bibliography #2-#11]   [End SCITE writing.]

I've also put the full MRD AI Essay, and the full listing of 11 annotated citations, in an open access cloud PDF file here:


  • I first mentioned SCITE in August 2023.
  • See JAMA 2024 on rules for use of AI assistance..
  • SCITE has full-text access to all public articles, plus licenses that allow it to present short quotes even from many firewalled articles.