\n
\nWho is this event intended for?: \;Statisticians and Scientists involved or interested in randomization techniques in pre-c linical and early clinical studies.
\n\n
\nWhat is the benefit of attending?: Hearing about modern rando mization methods.
\n
Cost
\nThis webinar is free to both Me mbers of PSI and Non-Members.
\nRegistration
\nTo register f or this event\, please click here
\nOverview
\nIn both\, clinical and preclinical studies ra ndomizing subjects to treatment groups is a key part of the study design. This step minimizes the risk of confounding. For large studies\, complete randomization is often enough to ensure that treatment groups are similar. However\, for small clinical trials and especially for in vivo studies th is approach is riskier. More elaborate randomization techniques can be app lied for those studies such as randomized block design. However\, the best results are achieved when algorithmic approach to subject allocation is t aken. One such approach is based on genetic algorithm that is rooted in Th eory of Evolution. The methodology minimizes the fitness function criteria for partitions of a dataset into balanced subgroups. The performance of t he algorithm was compared to random allocation and the exhaustive search u sing synthetic and real-world data. The results showed that experimental g roups created by this algorithm were more homogeneous compared those creat ed by exhaustive search. Additionally\, this algorithm is significantly le ss expensive computationally compared to the exhaustive search\, and the e fficiency gains increase rapidly as the number of subjects and design fact ors increase. We will discuss the theory behind this approach as well as t he extensions and variations of the algorithm such as simulated annealing.
\nSpeaker details
\n| \n Speaker \n | \n
\n Biography \n | \n
\n Abstract \n | \n
 \n Davit Sargsyan\, Johnson &\; Johnson | \n
Davit Sargsyan is an associate director in a no nclinical statistics group at J&\;J Innovative Medicine\, mainly suppor ting Immunology Discover R&\;D. He received an MS in statistics and com pleted his PhD in pharmaceutical sciences at Rutgers University. To date\, Davit has co-published over 70 scientific articles in peer-reviewed journ als on topics ranging from clinical outcomes of cardiovascular patients us ing a state-wide hospital admission data registry to natural compounds tes ting in in vivo and in vitro models. Davit&rsquo\;s research at Rutgers Sc hool of Pharmacy concentrated on computational methods and visualization o f omics data studying the effect of dietary phytochemicals on epigenome\, transcriptome and microbiome. | \n \; \n In both\, clinical and preclinical studies rando mizing subjects to treatment groups is a key part of the study design. Thi s step minimizes the risk of confounding. For large studies\, complete ran domization is often enough to ensure that treatment groups are similar. Ho wever\, for small clinical trials and especially for in vivo studies this approach is riskier. More elaborate randomization techniques can be applie d for those studies such as randomized block design. However\, the best re sults are achieved when algorithmic approach to subject allocation is take n. One such approach is based on genetic algorithm that is rooted in Theor y of Evolution. The methodology minimizes the fitness function criteria fo r partitions of a dataset into balanced subgroups. The performance of the algorithm was compared to random allocation and the exhaustive search usin g synthetic and real-world data. The results showed that experimental grou ps created by this algorithm were more homogeneous compared those created by exhaustive search. Additionally\, this algorithm is significantly less expensive computationally compared to the exhaustive search\, and the effi ciency gains increase rapidly as the number of subjects and design factors increase. We will discuss the theory behind this approach as well as the extensions and variations of the algorithm such as simulated annealing.\n |
END:VEVENT END:VCALENDAR