Friday, March 1, 2024

SCISPACE: An AI tool for biomedical paper research

Like SCITE, "SCISPACE" is a sort of PubMed on AI steroids.  You'll get AI-driven bells and whistles, compared to a plain-jane PubMed search.


On February 22, 2024, I highlighted SCITE, a tool for studying who's cited a key biomedical paper, how they've cited it, and seeing related papers. 

Here, a few words about another tool in this space, SCISPACE. Find it here:  [*]   

You can ask SCISPACE a biomedical question, and it comes back with five or six AI sentences with citations.  It also comes back with bibliographies of topical papers, and you can filture for features like "Top tier papers" or "Open access."

It does more, but that's a thumbnail sketch.


I asked it a topical question, "What is the evidence for the clinical utility of minimal residual disease (MRD) monitoring in cancer management?"

It came back instantly with this "Insight from Top 5 Papers."

  • Minimal residual disease (MRD) detection has shown clinical utility in cancer management. MRD monitoring is a powerful predictor of disease-free and overall survival, particularly in acute lymphoblastic leukemia (ALL) [1]. Different methods, such as polymerase chain reaction (PCR) and droplet digital PCR (ddPCR), have been used for MRD detection with high sensitivity and precision [2] [3]. MRD detection using ddPCR has been shown to detect micro-residual disease missed by PCR in one patient [4]. In non-small cell lung cancer (NSCLC), ultrasensitive fragmentomic assays using whole-genome sequencing (WGS) have demonstrated the ability to detect MRD and predict recurrence with high sensitivity [5]. In breast cancer, circulating tumor DNA (ctDNA) assays have allowed for the detection of molecular residual disease (MRD) prior to clinical evidence of recurrence, providing a lead time for intervention . MRD detection using ctDNA has shown promise in guiding therapy escalation in ongoing clinical trials . In colorectal cancer, tumor-informed ctDNA testing has been used to detect and quantify MRD, with higher rates of detection in liver metastasis . These findings highlight the clinical utility of MRD detection in cancer management, including early intervention and guiding treatment decisions.


It may not escape attention that the above instantly-created paragraph about MRD could be nearly cut-and-pasted for various purposes that are not super-demanding.  


In addition, SCISPACE came back with 10 papers by name, with links and PDFs if available.  

It also instantly provided a one sentence summary of each paper. Click below to see the top of this ten-row table.

Check YouTube for training and review videos for SciSpace or register with email for some entry level features.  More by subscription.


I asked the question again but specified I wanted results in lung cancer.

  • MRD detection has shown to be useful in lung cancer management. Studies have demonstrated that the traditional circulating tumor DNA mutation-based approach has limited performance in MRD detection, especially for early-stage cancer after surgery [1] [2]. However, a cfDNA fragmentomics-based method using whole-genome sequencing has shown great sensitivity in predicting prognosis and detecting patients with a high risk of recurrence [3] [4]. This method has been found to be more sensitive in predicting patient recurrence compared to the traditional circulating mutation approach [5] . Additionally, longitudinal MRD detection using circulating tumor DNA has shown high sensitivity and can potentially identify patients who may benefit from consolidation therapy . These findings highlight the potential of MRD detection in guiding adjuvant therapeutics and improving outcomes in lung cancer patients .

When I filtered in my query for "top tier papers," it came back in a couple seconds with an updated 10-paper list and again a write-up of first 5 papers.

  • Molecular residual disease (MRD) detection has shown promise in the management of lung cancer. Studies have demonstrated that circulating tumor DNA (ctDNA) analysis can be used to detect MRD in non-small cell lung cancer (NSCLC) patients after surgery or radiotherapy [1] [2]. Longitudinal surveillance of ctDNA has been found to be effective in predicting relapse and guiding adjuvant therapy decisions [3]. The sensitivity and specificity of ctDNA MRD detection have been shown to be high, making it a promising biomarker for relapse prediction [4]. Additionally, ctDNA analysis has been used to identify patients with MRD who may benefit from targeted therapies or chemotherapy [5]. These findings suggest that MRD detection using ctDNA analysis can play a valuable role in the management of lung cancer, allowing for early detection of relapse and personalized treatment strategies.


SCITE versus SCISPACE?  I haven't used each enough to give a masterful comparison.  But, on the surface, SCISPACE is designed to be driven by questions, as in the examples above, give you 10 citations, sort them by "open access" if you want, and give you a paragraph summary of the field driven by the top citations, as well as one-sentence summaries-at-a-glance of all the papers it pulled.

SCITE works better if you already know a key paper.  It gives you every paper that has cited it, which you could get elsewhere, but it shows you HOW each citing paper has cited your index paper.   That can be pretty darn powerful and it's  available in a second or two.  

SCITE pretty much depends on a subscription, as I recall.  SCISPACE has a free tier that is pretty good, but will cut you off after a limited number of queries.   The upgrade at SCISPACE is $12/mo but sold as $120/yr.   


Other resources of potential interest are Semantic Scholar, and Google Biomed Explorer. They're both free.
See also the general AI  sources and  The latter 2 are both general AI's, with a free tier, but which are especially adept with citations.  I showed an example of drafting a footnoted article with Perplexity in an earlier blog.

* At some point the address will be