Deep Learning for Health and Life Sciences 

Deep Learning for the Health and Life Sciences Valerio Maggio, PhD @l er i o m a g g i o v ma g g i o@a n a c o n d a .c o m 

m e p u n Who? B a c k g r o un d i n C S P h D i n Ma c h i n e L ea r n i n g R e s e a r c h : M L /D L f o r B i o Me d i c i n e P y t h o n G e e k 
 D e v e l o p e r A d v o c a te _a t _ An a c o n da a l s o m e 

Machine Learning “M a c h i n e l e a r n i n g i s t h e s ci e nc e (a nd a r t) o f p r o g ra m m i n g c o m p ut e rs so th e y c a n l e a rn fr o m d a t a” A u r él ie n G ér o n , Ha n ds -o n M a c h i n e L e a r n i n g w i t h Sc i ki t-L e a r n a nd T e n s o r Fl o w S o u r ce : bi t.l y /m l -s i m p l e -d ef i ni t i o n “(m l ) fo c us e s o n t e a c h i n g c o m p ut e rs h o w t o l e a r n w i t h o u t t h e n ee d t o b e p r o g r a m m e d f o r s p ec i f i c t a sk s” S. P a l & A . G u l l i , D e e p L e a r ni n g w it h K e ra s 

Machine Learning M a c h i n e l e a r n i n g t e a c h es ma c h i n es h o w t o c a rr y o u t t a s k s b y t h e m s e l ve s. I t i s t h a t s i mp l e . T h e c o m p l e x i t y c o m e s w it h t h e d et a il s L o u i s P ed r o C o e l h o , B u i l d i ng Ma c h i n e L ea r ni n g S y s t e m s w i t h P y t h o n (a n d th a t ’s p r o b a bl y o n e o f t h e re a so n w h y y o u ’r e h e r e :) 

The (Machine) Learning is about DATA D a t a a r e o ne o f th e m o s t i m p o r t a n t p a rt o f a m l s o l u ti o n I m p o rt a n c e: Da t a >> Mo d e l ? L e a r ni n g b y e xa m p l e s D a t a P r e p a ra t i o n i s c r uc i a l ! d a t a a l g o r i t h m s 

BioMedicine: another data case ? C o n t e m p o r a ry Li f e S c i e n c e i s a b o u t d a t a r ec e nt a d v a n c e s i n s e q u e n ci n g t ec h s a nd i n s t r u m e n t s (e.g. “b io -i m a g e s ”) h u g e d a ta s e t s g e ne r a t e d a t i n c r e d i bl e p a ce f ro m h u m a n o b se r v a t i o n t o d a ta a n a l y s is c h e m o -i nf o rm a ti c s (d r u g d i sc o v e r y ) R e s e a r c h I m p a ct —> S o c ia l a n d H u m a n I mp a c t 

Why Deep Learning, btw? S u b s et o f M L w / v e r y s pe c if i c m o d e l : (D e e p?) N e u ra l N e tw o r k s S ta t e o f t h e a r t T h e o r y ’50 / ’80 h w a cc e l e r a t i o n t o tr a i n (~n ew ) l e a rn i ng st r u c t ur e + c o m po s a b i l i t y (2018/23) 

Ml / Dl basics in a NutShell 

f e a t ur e s l a b e l s (r a w ) d a t a M L /D L 
 M o d e l T r a i n i n g T ra i ne d 
 M o d e l (u n s e e n ) d a t a T e s t P r e di c ti o n s Supervised learning s up e rv i si o n 

f e a t ur e s l a b e l s (r a w ) d a t a M L /D L 
 M o d e l T r a i n i n g T ra i ne d 
 M o d e l (u n s e e n ) d a t a T e s t S i m i l a r i t i e s /l i k e l i h o o d UnSupervised learning 

l a b e l s (r a w ) d a t a D L 
 M o d e l T r a i n i n g T ra i ne d 
 M o d e l (u n s e e n ) d a t a T e s t P r e di c ti o n s s up e rv i si o n f e a t ur e s Deep 
 Supervised learning 

What Deep Learning is about… 

Neural Networks A m u l t i-l a y e r f e e d-f o r w a r d n e ur a l n et w o r k c o mp o se d b y m u l t i pl e 
 d en s e (a.k.a. f u l l y -c o n n ec t ed ) h i d d e n l a y e r s a nd no n-l in e a r t r a n s f o rm a t i o ns 

No content

More details… 

More details… ReLu | sigmoid | tanh 

More details… Repeat for each layer… 

More details… Image Classification Task 

More details… S u m m a r y: 
 
 A (b a s i c ) D e e p N e u r a l N e t w o r k i s : • C o m p o s e d o f m u l t i p l e l a y e rs ; • E a c h L a y e r i s f (W x +b): • L in e a r m o d e l -> W x +b • f() is a n o n -l i n e a r f u n c t i o n G o a l o f t h e l ea r ni n g t a s k: fi n d i n g t h e va l ue s o f a l l 
 W s a n d bs (i.e. m o d e l p a r a m e te r s ) s o t h a t t h e c l a s s i f ic a ti o n e rr o r w o u l d b e m in i m u m. 

Supervised Learning Loop l a b e l s (r a w ) d a t a M o d e l P a r a m e t e r s L o s s l o s s p r e d ic t i o n s 

Supervised Training Loop breakdown & terminology (r a w ) D a t a - a .k.a. O b s e r v a t i o n s / I n p u t 
 I te m s a b o u t w h i c h w e w a n t t o p r e d i c t s o m e t h i n g. W e u s u a l l y w i l l d en o te o b s e r v a t i o n w i t h x . L a b e l s - a .k.a. T a r g e t s (i.e. G r o u n d T r u t h) 
 L a b el s c o rr e s p o n d i n g t o o b s e r v a t i o n s. T h e s e a r e u s u a l l y t h e t h i n g s b e i n g p re d ic t e d . F o l l o w i n g s t a n d a r d n o t a t i o n s i n M L /D L , w e w i l l u s e y t o r e f e r t o t h e s e . M o d e l f(x) = ˆy 
 A m a th e m a t i c a e x p r e s s i o n o r a f u n c t i o n t h a t t a k e s a n o b s e r v a t i o n x a n d p re d ic t s t h e v a l u e o f i t s t a r g e t l a b e l . P r e d i c t i o n s - a .k.a. Es t i m a t e s: V a l u e s o f t h e T a r g e t s g e n e r a t e d b y t h e m o d e l - u s u a l l y r e f e r r e d t o a s ˆy P a r a m e t e r s - a .k.a. W e i g h t s (i n D L t e r m i n o l o g y ) 
 P a r a m e t e r s o f t h e M o d e l . W e w i l l r e f e r t o t h e m u s i n g t h e w. L o s s F u n c t i o n L (y, ˆy): 
 F u n c t i o n t h a t c o m p a r e s h o w f a r o f f a p r e d i c t i o n i s f ro m i t s t a r g e t f o r o bs e r v a t i o n s i n t h e t r a i n i n g d a t a . T h e l o s s f u n c t i o n a s s i g n s a s c a l a r r e a l v a l u e c a l l e d t h e l o s s . T h e l o w e r t h e v a l u e o f t h e l o s s , t h e b e t t e r t h e m o d e l i s p r e d i c t i n g. T h e L o s s i s u s u a l l y r e f e r r e d t o a s L S o u r c e :D. R a o e t a l. - N a t ur a l L a n g u a g e P r o c e s s i n g w i th Py T o r c h , O ’R e i l l y 2019 

No content

Deep Learning in 
 Bioinformatics and BioMedicine A f e w e xa m p l e s 

Diabetic Retinopathy from fundus retinal images DeepDR 

DL for Digital Pathology Digital Pathology (Pipeline) Detect Tissue Region Transfer Learning 

DL for Digital Pathology Transfer Learning 

BioMedical Image Segmentation U-Net Note: AutoEncoder Model (Unsupervised Learning) 

Toxicity Prediction using Deep Learning ToX21 Dataset DeepTox Pipeline 

Now it’s time to switch to notebooks… 

Materials and Practical R e p o si t o r y: 
 h t tp s://g i t h ub .c o m /l er i o m a gg i o /d ee p-l ea r ni n g -c l a s s m y b i nd e r