Validation of Artificial Intelligence Containing Products Across the Regulated Healthcare Industries.

PURPOSE: The introduction of artificial intelligence / machine learning (AI/ML) products to the regulated fields of pharmaceutical research and development (R&D) and drug manufacture, and medical devices (MD) and in vitro diagnostics (IVD), poses new regulatory problems: a lack of a common terminology and understanding leads to confusion, delays and product failures. Validation as a key step in product development, common to each of these sectors including computerized systems and AI/ML development, offers an opportune point of comparison for aligning people and processes for cross-sectoral product development. METHODS: A comparative approach, built upon workshops and a subsequent written sequence of exchanges, is summarized in a look-up table suitable for mixed-teams work. RESULTS: 1. A bottom-up, definitions led, approach which leads to a distinction between broad vs narrow validation, and their relationship to regulatory regimes. 2. Common basis introduction to the primary methodologies for software validation, including AI-containing software validation. 3. Pharmaceutical drug development and MD/IVD-specific perspectives on compliant AI software development, as a basis for collaboration. CONCLUSIONS: Alignment of the terms and methodologies used in validation of software products containing artificial intelligence/machine learning (AI/ML) components across the regulated industries of human health is a vital first step in streamlining processes and improving workflows.

Adipose tissue Mest and Sfrp5 are concomitant with variations of adiposity among inbred mouse strains fed a non-obesogenic diet.

The expression of a subset of genes including mesoderm specific transcript (Mest), secreted frizzled-related protein 5 (Sfrp5) and bone morphogenetic protein 3 (Bmp3) in adipose tissue biopsies of C57BL/6J mice before exposure to an obesogenic diet were shown to be predictive for the development of obesity in mice after feeding a high fat diet for 8 weeks. This observation led to the supposition that adipose tissue expression of this subset of genes within inbred strains of mice could be associated with their susceptibility in the development of adiposity when fed a low fat diet. The analyses of male mice from 5 inbred strains showed average bodyweights ranging from 25.82 to 36.58 g at 16 weeks of age. Bodyweight was highest for AKR/J and adiposity correlated highly with bodyweight for all strains. Analyses of epididymal fat gene expression showed Mest, Sfrp5 and Bmp3 to be highly concomitant with adiposity across all strains of mice. Naked 1 (Nkd1), a gene previously shown to be associated with variations of adiposity in mice fed a high fat diet, but not predictive for the development of adiposity, showed no correlation with adiposity. In addition, the expression of Mest and Sfrp5 were tightly associated across the 5 mouse strains with the highest and lowest expression occurring in DBA/2J and C57BL/6J (B6) respectively suggesting a common mechanism for their regulation. Surprisingly, when independent cohorts for these 2 strains were fed high fat diet for 8 weeks, DBA/2J showed no further increase in Sfrp5 expression whereas expression levels for B6 mice were induced almost 20-fold. Analyses of (B6 x DBA2/J) F1 mice fed a low fat diet for 8 weeks showed intermediate levels of adiposity and gene expression for Sfrp5 and Mest suggesting a strong genetic basis for these differences.

Positional identification of TNFSF4, encoding OX40 ligand, as a gene that influences atherosclerosis susceptibility.

Ath1 is a quantitative trait locus on mouse chromosome 1 that renders C57BL/6 mice susceptible and C3H/He mice resistant to diet-induced atherosclerosis. The quantitative trait locus region encompasses 11 known genes, including Tnfsf4 (also called Ox40l or Cd134l), which encodes OX40 ligand. Here we report that mice with targeted mutations of Tnfsf4 had significantly (P

A torrid zone on mouse chromosome 1 containing a cluster of recombinational hotspots.

Within the 2.38-Mb Ath1 region of mouse chromosome 1, 42 of 45 genetic crossovers from crosses between C57BL/6J (B6) and either C3H/HeJ (H) or Mus spretus (SPRET) occurred in four zones (A-D); zone A, 100 kb long, contained a cluster of at least four recombination hotspots. F1 sperm assays indicate that within this "torrid zone" the most active hotspot (A3) can initiate recombination on H and SPRET but not B6 chromosomes. The A3 DNA sequence contains a (G/C)TTT repeat, long stretches of A or T, and a cyclic variation in AT content. Recombination was drastically reduced in a cross between B6 and a B6.SPRET Ath1 congenic strain, but was unaffected in a B6 x B6.H Ath1 congenic cross. Similar nonrandom clustering of hotspots has been observed in yeast and the major histocompatibility complexes of human and mouse. To the extent that torrid zones are a general feature of mammalian genomes, they have considerable implications for genetic mapping strategies in both human populations and mouse crosses.

Confirmation and high resolution mapping of an atherosclerosis susceptibility gene in mice on Chromosome 1.

Previously, we demonstrated that Ath1 is a quantitative trait locus for aortic fatty streak formation, located on Chromosome (chr) 1, with susceptibility in C57BL/6J mice and resistance in C3H/HeJ and BALB/cJ mice fed an atherogenic diet. In this study, we find an atherosclerosis susceptibility locus in the same region of Chr 1 by constructing two congenic strains with the resistance phenotype transferred from different resistant strains, PERA/EiJ or SPRETUS/EiJ. By backcrossing one congenic strain to C57BL/6J and testing recombinant animals, we reduced the distance of the atherosclerosis susceptibility region to 2.3 cM between D1Mit14 and D1Mit10. Further testing of nine recombinant animals showed that eight of the nine were consistent with a further narrowing between D1Mit159 and D1Mit398 a distance of 0.66 cM. This region encompasses a number of potential candidate genes including the thiol-specific antioxidant gene Aop2, also known as peroxiredoxin 5 (Prdx5). AOP2 is capable of reducing hydroperoxides and lipid peroxides in the cell. To investigate Aop2 as a potential candidate, we mapped Aop2 in our backcross and localized it to the atherosclerosis susceptibility interval. We determined that Aop2 is highly expressed in atherosclerosis-related tissues including liver and heart. We also found an inverse correlation between Aop2 mRNA in liver and atherosclerosis phenotype for strains C57BL/6 and the resistant congenic derived from SPRETUS/EiJ. Since LDL oxidation has been implicated in the pathogenesis of this disease, and AOP2 possesses antioxidant activity, we suggest the role of Aop2 in atherosclerosis susceptibility needs to be further explored.